I think that the scores under 1% are rumours more than other.
23andme use different systems than Dodecad.
Yes, I know they're different systems; the question is which is more accurate.
I'd put my money on 23andme any day.
I also don't know what you mean by scores under 1% for Sicilians on 23andme being "rumours". They're not rumors, they're my shares.
They may or may not be representative, but they're "real" Sicilians.
These are all from Messina, which I admit might impact the scores given that the mountains near there were relatively autonomous from Muslim rule.
1) 2.4% NAfrican
0% SSA
2.) 2.2% NAfrican
.6% SSA (.3 W. African, .2 East African, .1 Broadly SSA)
3.) 4.1% NAfrican
.3% West African
4) .1% NAfrican
.1% SSA
This Palermo sample also has very low numbers, although there's one Calabrian grandparent in this case.
.2% NAfrican
.1% WAfrican
By the way, is there any sort of reciprocal arrangement so that people who tested with FTDNA can run their results through the 23andme algorithm. I'm not talking about running their results through the gedmatch versions of the various hobbyist calculators, but actually through the 23andme algorithm. I know there's a defunct company that did have that arrangement with 23andme, just don't know if one has been worked out with FTDNA or Ancestry.
Ed. I don't share with anyone from the southern coast where there was more documented actual Berber settlement, so those scores may be very different.
I think for accurate broad autosomal ancestry information you don't need very many samples, but for this kind of very minority ancestry you might see swings from region to region or even town to town, especially given how little movement there was from some of these more isolated rural areas. It's the same situation that you see in some Alpine isolates in Italy. They're not representative of the population as a whole.