Even though it would be nice if the software somehow recognized which faces were more important than others, I don’t think it matters.
As long as the software knows to keep touch points away from the edge of any/all faces, I believe it should solve the problem.
Even if a touchpoint ends up getting generated on a face one doesn’t want any touch points on, that is still way fewer auto-generated touch points that require editing/adjusting compared to how it is now.
Also, since almost all touch points are applied to the face(s) that are facing towards the build platform, the orientation one places their part already dictates which face(s) will receive the touch points. So as long as the auto-generated supports stay away from the edge of the faces that they were anyway going to be applied to, then that should work as a good set of support structures without all the labor of moving them!
I’m sure there may be other considerations, but I do believe it is doable.