According to the Kriesel III majority, the "primary justification" for keeping samples is that "match confirmation" ensures "the continued accuracy and integrity of the CODIS system." That is:
Upon receiving a CODIS Match Report, the ... FDDU [Federal DNA Database Unit] locates the retained blood sample ... , re-extracts junk DNA from it, and runs a new analysis. If the newly generated profile is the same as the one in CODIS that formed a Candidate Match, the match is confirmed and the accuracy of the match between the CODIS profile and the identified offender is ensured. The confirmation of the CODIS match is thus achieved by comparing the profile generated from the retained sample with the Codis profile. Although CODIS has not yet encountered such a “mismatch” or “misidentification” error, in the event that the generated profile did not match the CODIS profile, the lab would then determine what caused the error and, presumably, prevent similar errors from occurring in the future.Let's pause to note the infelicities here. First, a cavil about phrasing. The government does not just extract "junk DNA" (a term that is best avoided). It extracts the entirety of the physical genome (all the DNA) but types only a small number of loci that do not reveal much about the individual's health or fitness. Second, a match to the DNA sample that is on file (as a spot on a card) does not "ensure" that the "identified offender" matches the crime-scene sample. If the name associated with the card is incorrect (Mr. Jones's DNA has been recorded as Mr. Smith's), the profile is not the offender's even though both DNA profiles match. 
Still, the retyping step does protect against recording a profile in the database that is not the one on the card; moreover, as Judge Schroeder observed, "[i]t also enables pre-arrest confirmation of a match."
The dissent's response is that
The government's argument is severely undercut by the fact that the CODIS database has never led to a false identification of a suspect ... . Moreover, if by some remote chance, there was an unprecedented error in the CODIS system, that error would be swiftly discovered when the CODIS-identified suspect had a new blood sample drawn and the new sample was compared with the DNA found at the crime scene, as is the regular practice. For this reason alone, the government's rationale is wholly theoretical at best and, to put it bluntly, is entirely without merit.Entirely without merit? Surely a DNA database can create an ordeal for a suspect who is arrested even if later testing of an entirely new sample from him leads to his release.  Judge Reinhardt seems to maintain that a mistyped DNA sample or a misrecorded profile could not "infringe the liberty of the misidentified suspect" because it merely "pertains to not taking a blood sample from the wrong person as a result of a CODIS misidentification." But police do not simply ask suspects who come to their attention because of a database hit to mail in a DNA sample at their convenience. Often, they arrest these suspects. Although the jailed suspects may have a get-out-jail-free card in the form of their DNA, it can take time to play it. 
The stronger argument is that the indignity and injury that comes from being an active suspect in a criminal investigation and, quite possibly, being detained and interrogated is, as the dissent ultimately recognizes, "an unfortunate occurrence, but avoiding this wholly theoretical and comparatively minor infringement on the suspect's rights cannot, by any measure, justify the retention of the entirety of that individual's, and millions of others', private genetic information for the rest of their lives."
But even this formulation is misconceived. The pertinent balance is either (1) between a single individual's risk of being falsely arrested as opposed to the same individual's risk of having truly private areas of his genome examined by the government, or (2) between these same risks summed over everyone in the database. To balance between a single individual's risk of being falsely identified and everyone's risk of improper acquisition or disclosure of genetic data is to place a very heavy thumb on the scales.
The dissent also dismisses the government's argument that avoiding public disclosure of laboratory or clerical errors is a good reason to hang on to samples. This point is well taken. Why should the government fear that the public might discover the obvious--that some mistakes are possible when an organization maintains millions of records? CODIS's efficacy (and public confidence in it) would not suffer if preliminary false matches are rare and inevitably corrected by confirmatory testing of fresh samples, as the government claims.
Finally, the dissent suggests that the FBI's "quality assurance" strategy of "randomly re-testing 1% of samples that were received by the FBI laboratory in the previous six months" cannot justify retaining millions of samples for more than six months. Certainly, less drastic methods of quality assurance are available. For example, the FBI could keep only 1% of the samples for re-testing. But this procedure would make it more difficult to correct a large number of profiles if they turned out to be incorrect. Suppose that a discrepancy resulted from a problem that affected all the samples placed on a tray with many wells on a given day. (This may not be the current technology, but it illustrates the broader point.) It might be easier to retrieve these samples from storage and retype them than to locate and collect more DNA from the individuals who provided the affected samples. However, if errors are as improbable as the dissent maintains, this precaution would not be justified.
Interestingly, one traditional argument for sample retention is absent from both opinions (apparently because the government did not raise it). What if some of the old loci are retired and new ones are used as replacements? (Maybe the former will be found to have privacy-laden associations with phenotypes?) Sample retention obviates any possible need to obtain new samples--a task that could prove quite onerous given the size of modern databases. Then too, what if outer-directed database trawling  is undertaken? A recent report proposes that "familial searching" as implemented in California could classify crime-scene samples from not-so-close relatives as coming from first-degree ones, causing turmoil for the first-degree relatives and complicating matters for the police.  One solution (if this is a real problem) would be to analyze more loci in the samples to ascertain the relationship with greater confidence. [9, 10]
In sum, Judge Reinhardt's complaint that "tens if not hundreds of millions of dollars expended on maintaining a totally unnecessary and wholly pointless system of collecting and maintaining tens of millions of blood samples indefinitely in a national warehouse" seems overdrawn--but the justifications discussed in the Kriesel III opinions for indefinite and widespread sample retention also seem strained. The value--immediate and potential--of sample retention is more complex than the case suggests, but it is unlikely that the government would suffer greatly if it had to destroy samples from individuals like Kriesel, who have completed their sentences. Indeed, however one comes out on the issue of the individual's Fourth Amendment right to compel eventual destruction or disgorgement of databank samples, even initial sample retention may not be the best public policy. 
1. Ninth Circuit Upholds Indefinite Retention of DNA Samples: The Majority Opinion in Kriesel III, July 17, 2013, http://for-sci-law-now.blogspot.com/2013/07/ninth-circuit-upholds-indefinite.html
2. Ninth Circuit Upholds Indefinite Retention of DNA Samples: The Dissent’s Perception of the Loss of Privacy in Kriesel III, July 18, 2013, http://for-sci-law-now.blogspot.com/2013/07/ninth-circuit-upholds-indefinite_18.html
3. Ninth Circuit Upholds Indefinite Retention of DNA Samples: More Problems with Judge Reinhardt’s Dissenting Opinion, July 25, 2013
4. Police Finger Wrong Man after DNA Data Mix-up, Asahi Shimbun, Mar. 22, 2010, http://www.asahi.com/english/TKY201003210142.html
5. Linda Geddes, DNA Super-network Increases Risk of Mix-ups, New Scientist, Sept. 5, 2011, http://www.newscientist.com/article/mg21128285.500-dna-supernetwork-increases-risk-of-mixups.html#.UhJ-zH-wXdU
6. Jack Doyle, Innocent Man Spent Five Months in Prison After Forensics Mix-up Meant He Was Falsely Accused of Rape, Daily Mail, Oct. 1, 2012, http://www.dailymail.co.uk/news/article-2211365/Adam-Scott-Innocent-man-spent-FIVE-MONTHS-prison-forensics-mix-meant-falsely-accused-rape.html#ixzz2cSKkGvBI
7. David H. Kaye, The Genealogy Detectives: A Constitutional Analysis of “Familial Searching”, 51 Am. Crim. L. Rev. 109 (2013), available at http://ssrn.com/abstract=2043091
8. Rori V. Rohlfs, Erin Murphy, Yun S. Song, Montgomery Slatkin, The Influence of Relatives on the Efficiency and Error Rate of Familial Searching, 8 PLoS ONE e70495, http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0070495
9. Jianye Ge et al., Haplotype Block: A New Type of Forensic DNA Markers, 6 Forensic Sci. Int’l Genetics 322 (2011)
10. Chad Huff et al., Maximum-likelihood Estimation of Recent Shared Ancestry (ERSA), 21 Genome Research 768 (2011)
11. David H. Kaye, Behavioral Genetics Research and Criminal DNA Databanks, 69 Law & Contemporary Problems 259 (2006), available at http://ssrn.com/abstract=1411861