I modeled the Fulanis on the datasheet. In "North-African," I used the samples, Gaunche (Berber-related people from the Canary Islands who are extinct), Saharawi, and Taforalt (very ancient, 15,100-13,900 BP). On the Gambia_WOLLOF, I used two samplings from a set of samples that approximated the internal components closest to the collective average, which also had the best fits (vetted through another mixture modeling with Sudanese, Yoruba, and Taforalt) to minimize extra intricacies.
Here I replaced the Gambian Wolof samples with Yoruba. You see that the "North-African" increased in percentage. It's because those Wolof samples are not entirely West African. About ~15 of their DNA is attributed to Nilo-Saharan-related and ancient North African ancestry, which absorbs or masks North African DNA into Wolof, and that's why it is increased when using Yoruba, a relatively non-admixed sample.
And here is one including Sudanese.