In the US it does happen - in legal terms it's called a "pure cold hit." The legal problem with such cases is that the DNA "match" is never a 100% match, it's a statistical statement called a random match probability - such as, there's a "1 in 1.1 million chance that this sampling of genetic material at this number of locations in the genome could match a random, totally innocent person." And sometimes that random match probability is much weaker, like "1 in 450." Some pure cold hit suspects have ended up being exonerated because it was discovered - either in the investigatory stage or on appeal - that the cold hit was coincidental or erroneous. And this is actually only going to happen much more often, because criminal DNA databases like CODIS are getting larger by the day (due to laws about taking DNA from felony arrestees and other detained people) so there's ever larger likelihood that coincidental matches will occur. This article does a great job explaining this:
The Dark Side of DNA Databases
And then there's this:
DNA's Dirty Little Secret - Type Investigations
That article talks about the case of John Puckett. In 1972 a nurse was raped and murdered in San Francisco. The initial suspect was a man named Robert Baker, who had escaped from an asylum a month prior, had raped a woman less than a quarter-mile from the murder victim's home, and who, when questioned, had a parking ticket in his van that had blood on it matching the nurse's blood type. But the case ended up going cold until DNA technology caught up. In 2004 authorities re-opened that case and compared the deteriorated sample of DNA found on the victim's body to California's state database version of CODIS (which contained over 300,000 DNA profiles) and found one cold hit match, to a then-71 year old man named John Puckett, who shared 9 alleles with the DNA profile found on the victim's body. At his trial, the primary evidence against him was the DNA profile match - that a person picked at random would have a 1 in 1.1 million chance of matching the DNA found (as well as the fact that he had lived in the general Bay Area around the time of the murder and had a previous sexual assault conviction). The judge allowed the jury to hear the defense's argument that other than the cold hit, there was really no other evidence against John Puckett, but did not allow the defense to tell the jury that the chance of sharing 9 alleles is not uncommon and that in fact there were probably about 40 other people in California who matched the sample as exactly as he did. John Puckett was convicted of this crime and appealed.
The bottom line, and why all of this relates to Delphi, is this:
1. If the Delphi investigators have a deteriorated or partial sample, OR
2. If it is a sample with mixed contributors (including the victim)
THEN even a partial match in a DNA database may not be very strong evidence because it could "include" quite a lot of innocent people.
ETA: If I had to guess, I'd say the Delphi investigators have #2 - a mixed sample that contains, say, Libby's DNA, Kelsi's DNA, and the (perhaps partial) DNA of an unknown male contributor. So LE doesn't know if the unknown profile they've isolated belongs to the killer or even if they've isolated it correctly.