Again, you're overstating it. There aren't a lot of such people They're at best a tiny minority and usually recent and know about the admixture like my cousin. I've been in the pop gen game for like 15 years and have never noticed some notable commonness of Arabian admixture in coastal northerners that isn't already in all other Somalis.
If you're referencing something like how the Yemeni in Somalis using nMonte and G25 can vary, even in interior Somalis, that's not really "Yemeni admixture variance", saaxiib:
To be honest, I think people are wasting their time trying to use G25 for modeling individuals. The PCA gets exceedingly wonky for many samples in this setting probably because its dimensions are now too affected by the sheer number of samples affecting the formation of the principal components.
Hence why you'll get nonsensical stuff like Somali samples whose overall MENA ancestry is negligible in range across the ethnic group (2%) then nonsensical results like one sample magically being 10% Arabian and the other 15%. It makes no sense until you understand how PCAs can work and what can affect their output; namely the number of PCs you limit the output to, the SNP count of the samples, and of course the samples used themselves. Not to mention how easy it is, like with ADMIX for things to be skewed by drift even within individuals samples.
When you consider those points it starts to make sense how an ethnic group that clusters so tightly, forms its own ADMIX component, has the same mtDNA lineages across a wide area and is dominated by mainly two greater Y-HGs is showing variation in G25 that makes zero sense. If we seriously had diverse Mota and Arabian scores we should cluster as heterogeneously as Oromos but we don't so it's clearly the methodology that's shooting out incorrect results. Not to mention that you can actually make almost any of these Somali samples show these ancestries they're supposedly missing if you just structure modeling a certain way. Folks need to exercise some critical thinking...
I really would advise people to just stick to the averages when studying ethnic groups via G25 and hope someone creates something like a really fine-scale ADMIX run or something similar for modeling individuals. This guy probably doesn't have any really different amounts of Mota or Arabian than anyone else given his typical overall admixture levels...
If you're fitting within the ~44-46% MENA admixture range of all other Somalis and can probably form the "Ethio-Somali" cluster with them or fit 95-100% within the Somali cluster on 23andme then, trust me, the variance in Yemeni nMonte is showing you is
nothing.
If you actually show notable MENA on something like 23andme and/or vary from other Somalis notably in terms of admixture levels then have elevated elevated Yemeni that doesn't even look al-Jawf using nMonte like with all other Horners and you know about recent admixture then we can talk and that almost never happens.