Everything depends on 23andme's reference samples, which, other than the few academic samples they use, are not scientifically chosen, and therefore might not be representative of the areas involved. Many more testees at 23andme are Irish than English, because most clients are from the U.S. and there's a big population here that identifies as Irish. Any people of English descent usually consider themselves just "American".
In Italy, for example, the dot looks to be between Rome and Campania, which makes perfect sense as most of their testees are from southern Italy, but they have the big block of Tuscans from the 1000 genomes project which pulls the centrum northward. That dot
does not mean that's where the most genetically "Italian" people can be found.
As for the "French-German" cluster, that is one of the least reliable of any of the clusters according to 23andme's own analysis. What they basically did is use the French Lyon sample, which is the only one that was available for a long time, and the Swiss samples. The southern French sample now available was
not included. A high number for this basically tells you how "Swiss like" you are, because there are a lot more samples for that. Sardinian isn't much good either.
Scroll down to Testing and Validation at the following link:
https://www.23andme.com/ancestry-composition-guide/
The "map" results are totally predictable if you know how 23andme works.
By misinterpreting these results novices can be led astray.
Oh, in searching for the link for this white paper I discovered that in more recent white papers 23andme is telling people their results go back 500 years
and more. Just google 23andme white papers and you'll find it
. Of course, they're not going to acknowledge how often they were told that, including by me.