I don't think you should read too much into a statistical matching of your DNA markers with statistical samplings of populations in which individual samples also vary widely. It's the modern version of those occupational survey systems (Kuder?) that really just tell you what population your answers are most similar to, with very little useful insight.