Wrong on both counts.

I've looked at the shadowdemo.c from the shader workshops and it uses the stage feature to make the depth view render before the camera view. You could make a larger chain to render depth -> camera -> combine. Combine is basicly a postprocessing effect with 2 input bitmaps.

If you don't set camera.material, no material is overwritte. So just don't set it, and set the appropriate shader for every entity individually. For example, a normal object you give the normal shadow shader, the normalmapped objects are given the normalmap+shadow shader, etc.

Creating a normalmap+shadow shader requires some shader knowledge, but is not very hard. If you take the time to follow the shader workshops, I'm sure you can do it.