A small note on the DNA source data and the alleles/stutter issue:
Here's a (admittedly over-simplistic) analogy to help people better understand what's going on here:
In the UK national lottery's main game, one has to choose 6 numbers between 1 and 49 inclusive. The odds of all your six numbers being the ones the machine picks in any given week (and thus winning the jackpot) are around 1 in 14 million.
So, imagine that one week the machine selects the following numbers: 7, 12, 27, 29, 34, 41. Imagine that you have a lottery ticket in front of you. Imagine that written on that lottery ticket are not six numbers, but eighteen numbers. Next, imagine that (even though this is clearly impossible in the real world of lottery tickets), these 18 numbers may be "false" numbers, or may be the combination of as many as three different tickets.
So, you see that there are 18 numbers on the ticket in front of you. They are 4, 5, 7, 10, 12, 14, 15, 16, 20, 23, 27, 29, 30, 32, 33, 34, 41, 43.
Now you look back at the newspaper, which has printed the winning numbers for this week (7, 12, 27, 29, 34, 41). You now decide that 4, 5, 10, 14, 15, 16, 20, 23, 29, 30, 32, 33, and 43 are actually "false" numbers, and that 7, 12, 27, 29, 34, 41 are "real" numbers" - and you don't explain how you distinguish between "real" and "false" numbers. You call the lottery operator, and excitedly tell them that you have a winning ticket.
Only, of course you don't - unless you can show definitively that there is very good reasoning for distinguishing between the "real" and "false" numbers on the ticket.
What's more, if some or all of the "false" numbers are now added back in, you can't even say that if the ticket is a mixture of three different tickets, one of these tickets must be the winning ticket (7, 12, 27, 29, 34, 41). For example, the "7" might have been part of a different group of six numbers (for example including 30 and 43). So you can't even say that the winning ticket was a contributor to the overall ticket in front of you.
And this - in a very simplified analogy - is the issue with the DNA on the bra clasp. Stefanoni has arbitrarily chosen which peaks to classify as alleles, and which to dismiss as stutter/noise - and there appears to be no scientific rationale behind her selection methodology. And - surprise, surprise - the peaks she chooses to count as alleles all correspond to the "winning ticket" of Sollecito's alleles.
And the analogy extends further: if we re-admit some or all of the other peaks on the graph, then we are left with the conclusion that Sollecito's y-haplotype might be present, but equally the peaks ascribed to him might also match with other peaks in some other individual's profile.
(PS: I realise that this is far from a perfect analogy, but still think that it has validity in helping to explain the problems with Stefanoni's interpretation of the bra clasp DNA)