anglolawyer,
One possible problem with his approach is that it might be suspect-centered, perhaps depending on how it is applied. I think a great deal turns on the question of how the hypotheses he uses are generated. It would be helpful to know more about how the forensic science community views his work in this regard. Ultimately, we will have to wait and see how generally his method of mixture analysis is adopted.
As I understand it, a simplified explanation of mixed-DNA analysis and the problems related to suspect-centric evaluation might be the following:
Suppose a sample is analysed, and it shows peaks at the following numbers:
1, 2, 3, 4, 5, 6, 7, 8, 9
Suppose, just for the sake of this example, that every human being had peaks at four (and only four) numbers. So (for example) I might be 2, 4, 7, 8. You might be 1, 3, 4, 9. And so on.....
Now, if the person conducting the test is asked to make a tentative evaluation of whose DNA is on the sample - but without being given any reference DNA profiles to work with - it's very possible (probable even) that they would say something along the lines of: "It's difficult to make any positive singular matches, since there's clearly such an admixture of profiles that singling out any one contributor would be next to impossible.
But suppose now that the person conducting the test was given the reference sample of a particular individual, together (perhaps) with the information that this individual was a suspect. Suppose that this reference sample was 3, 5, 6, 9.
There is now a very real possibility that the person doing the test might look again at the tested sample, and - consciously or unconsciously - find ways to "match" the suspect to the sample. In this example, the tester might find ways to rationalise that the peaks at 1, 2, 4, 7 and 8 were stutters or other forms of artifact, and that therefore the only genuine contributory peaks were at 3, 5, 6 and 9. And hey presto! Suddenly the tester is able to conclude that the tested sample "matches" the profile of the suspect.
I think I have the driving principles of this issue correct - but anyone should feel free to correct me if I've got any aspect fundamentally wrong. Again, my example is meant to greatly simplify the real-world genetics, in order to highlight the underlying principles.