Arabians and North Africans are related, so how to precisely measure that stuff I don't really have a strong opinion on.
But I think the Somali type/Eastern Horn HG is much higher (8-12%) than what tests show using Mota as a reference who may not have been close to them. Just look at the stark difference between the Kenya LSA samples and Mota. HGs have much bigger genetic distances than agricultural groups do.
I hear they are going to release 18,000 year old SEA HG genomes soon, which may bolster this theory of mine.
Yemenis are easier to sus out because of their excess Iran-Chalcolithic. Just run any model with the current earliest Cushites + Yemenis (or Arabians like Saudis or BedouinB) + Mota or another East African HG and see the results for yourself. Or use Natufians and Dinkas in place of the Cushites. Somalis always show 5-10% Arabian based on the model whereas Xabashis show 20-30%.