The problem with the DNA tests we use today, assuming even perfect cases with zero contamination or mistakes (whole other argument there), is that (in my non-expert understanding):

    * Taking the samples and using several solutions to snip them at given patterns.
    * Taking another solution and promoting replication of the fragments.
    * Using some dye and a weight sorting channel to bin the fragments by weight.

This results in a kind of 'bar code' that graphs the distribution of weights as a VERY crude hash of samples of DNA.

It's pretty useful for determining things like IF it is LIKELY that individuals are related.

It's also pretty good at confirming negatives (We're sure someone is NOT related to X).

It's not that great at confirming positives. That is, the results are both subjective and ambiguous given the hashing. A "positive" result here is really more of an "OK, it's likely we should run the real and expensive check, evaluate if someone might be a suspect by other merits, etc."

I would, offhand, consider a "positive" above to be enough evidence to produce /suspicion/ and /warrants/ to locate other specific evidence to ascertain an actual guilt or innocence based on harder evidence. If an actual "sequence the whole set of samples" option enters the realm of feasible tests then it would also warrant actually doing that.

Come to think of it, I'm not sure how the Ancestry/etc novelty DNA tests work. I'd assume they've isolated a few specific markers they're looking for and the processes are optimized towards identifying those and comparing combinations for those specific traits.