The fact that the TSI sample is much larger (102) than the HGDP sample(8) does not necessarily mean that it is more representative of Tuscany as a whole, as the entire TSI sample was taken from one small village outside Florence, and genetic variation does exist within Tuscany. For example, the HGDP sample, (which shows .3% North African) lists the Mediterranean component as 47.8% versus 37.6% in the TSI sample. There are other differences as well.
The 25 person Zing sample is only based on 13,000 snps so I understand why that would not be factored into the computation. The Henn et al study, however, uses 195,000 snps, which should give a certain amount of confidence. That provides a figure of l.l % for north west African, and again 0 for the sub-Saharan and East African groups.
In the case of Tuscany then, an approximately 1% figure for North West African seems reasonable.
However, more generally, I am uncomfortable with a methodology for the making of maps (which are then widely distributed) which picks which studies to include strictly based on sample size. Indeed, I think that barring some generally acknowledged problem with a particular academic study, no such results should be excluded. For me, it calls into question the reliability of the maps in general.
I