"DNA matches aren't always a lock" is the headline of this LA Times story by Jason Felch and Maura Dolan. Cases where a suspect is identified by a "cold hit" database search involve different probability considerations than the typical case where a person already a suspect based on other evidence is confirmed as the perpetrator by a DNA test.
By analogy, if you buy a lottery ticket for a drawing, the chances that your ticket will win are extremely small. However, the chances that somebody will win from that drawing are often pretty good. If the chance that a person would match a profile if selected at random are one in a million and the database searched for that profile has millions of people, then there is a decent chance that an innocent person will be a random match. Of course, with fresh samples and modern techniques, the random match probability is 1/quadrillions, not millions, so the chance of random false positive remains very small. With old, degraded samples and only a few markers available, though, the chance of a false positive match is significant.
Is this story accurate? I would be interested in hearing from DNA experts.