Hi, Pascale-
I don't see what the issue would be with aligning EPI->T1w_with_skull to get Transform_A, and then T1w_without_skull-> other image to get Transform_B. Then you can get EPI->other image via concatenating Transform_A and Transform_B. This works because the T1w space does not change depending on having the skull or not. And you will want to concatenate the transforms, anyways, because applying each separately to the EPI would lead to unnecessary and unwanted smoothing.
Does that seem reasonable?
--pt