Fiorito et al doesn't have enough West Asian populations: there are no Iranians, Kurds, no Caucasians like the Adyghe, Lezghins, Abkhasians, no Iraqis etc. The addition or lack of certain populations makes a big difference and significantly changes where certain groups land in relationship to the others, as does which samples are used from which areas and therefore which data set is used.
Lazaridis et al make that point in their supplement. There's a whole section starting on page 76 entitled "Why do our PCAs correlate so poorly to geographic maps of Europe?" They could have added, Why did Novembre et al and some other studies produce PCAs that match that map while we don't? It seems to be that the PCAs that do match the map were made using Popres samples. Those that don't use them are different, including Lazaridis, Skoglund, and Behar.
After an extensive analysis, " We conclude that the pattern of two discontinuous clines that we observe is not an artifact of genotyping errors in the Human Origins data set. At the same time, these analyses highlight how sensitive PCA is to the specific geographic distributions of samples used. Thus, while PCA can be used to suggest interesting hypotheses about history, PCA results are not always unambiguously interpretable in terms of history, and need to be complemented by other types of analyses to produce convincing inferences. In this context, it is crucial that the cline of ancestry proportions that we measure in this study (S114 and S117) is inferred based on analyses that do not depend on the relatiave representation of different regions (as does PCA)."
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4170574/
If I had to go with one PCA I'd go with either the Lazaridis one or the Behar one.
That's why grand theories or pronouncements can't, in my opinion, be drawn simply from one PCA, and why I felt it important to highlight, in post # 100 the different perspective provided by the FST numbers, and also to mention that certain samples might not be very representative of a country or even a region as a whole. Then, as I also mentioned in that post, PCAs don't capture very much of the total genetic variation.