It is certainly not dumb, but makes sense to c_trace only a certain entity. But can't you use the existing features for this - for instance, setting all entities to group 1 and the entity to be detected to group 2?

For checking if a position is inside a certain mesh, I think the easiest way is a c_trace from the camera to the position and back from the position to the camera. If the mesh is hit in the first case but not in the second, then the position is inside the mesh - at least that's the theory.