I would venture to suggest that it is the *samples* which do this, not the database - am I right in thinking that when they narrow down their list of suspects the samples are re-processed to a higher accuracy than the profiles on the database? Or at the very least to confirm that the database isn't wrong?
Not so far as I am aware. The system I was familiar with looked at a particular sequence along the DNA strand and, to confirm a match necessitated a string of ten identical "bases" (adenine, thymine, cytosine and guanine), plus a sex identifier. It had been calculated that there is only one chance in a billion of two people having the identical profile. Nearly 19 out of every 20 people in the UK will have a DNA profile which is unique in the country (including the 90% of the population who are not on the database). Even where a duplicate does occur, the odds that an identical profile is found at the scene of a crime which matches a local
individual with a criminal past
must be infinitesimally small. In spite of the minuscule odds of a false match, if someone insists that the DNA could not be theirs, the police will have to produce corroborative evidence in addition to the DNA. Mostly, that's not too much of a problem.