I just happened to see this. If this algorithm is basically the same as FTDNAs from the period when Razib Khan set up the program, then the "Italian" reference populations are the HGDP, Hap Map, and Estonian Bio-Center samples, to the best of my recollection.
Within the latter there's a southern Italian population, but the majority of the reference sample is going to be low on southern Italians. If they also just dumped in all the Tuscan samples, which far outnumber those of the other groups, then the centrum is going to be pretty close to Tuscan, which makes some sense.
What I think always gets lost in these discussions is the fact that the results always depend on where the centrum is located, and that's to some extent a subjective decision. One of these companies had a "Balkan" cluster which covered northern Italy. So, northern Italians came out with huge percentages of "Balkan" ancestry. If they had called that cluster "North Italian", which is just as plausible, then people in the northern Balkans would have had huge percentages of "Italian". If a company creates a "Greek" cluster, then lots of southern Italians are going to get huge percentages of "Greek" because contrary to the fantasies of people like Sikeliot, these are very similar populations. When you don't have a specifically Greek centered cluster, then Greeks are going to get a lot of Italian, depending on where the Italian cluster is placed.
This is to some extent a problem for every ethnicity, but because Italy has so much more variation, especially in the north, it's more of a problem for Italians unless they understand how these things work. Just think of it this way: there's less than two dozen samples from northern Italy, which has a lot of variation. Let's say that in their reference sample there is no academic sample from the Veneto, as I'm pretty sure there isn't. There just aren't going to be matches in a certain percentage of the alleles, so it will look for related alleles in another "cluster". If they had a Veneto sample, the "Italian" percentage in people from the Veneto would go up.
It's just how these things work. There's no "right" percentage if you're talking about percentages of historical, recent populations.