One of the problems is that there are so many DNA samples around that there are starting to be false positives on family relations. The DNA class I took a while back said that the odds are one in a million that a stranger off the street would share 12 of the areas they look at for identification--but that means in an area the size of Greater Boston, with a population around 7 million, there are 6 or 7 other people out there who could be potentially identified as relatives. In areas with a more homogeneous gene pool, such as the upper midwest where so many people are of northern European ancestry, the rate goes even higher. And most of those people will not in fact be relatives, or rather are related so far back up the family tree that it's meaningless.
So the odds pretty good that she's related to someone in the criminal database, but it would require hundreds of hours of followup to determine whether anybody is actually related to her and they are likely saving it for a last resort sort of thing.
p.s. I assumed you meant a family-relations search of the state prison inmate databases. For Codis, they have better techniques.