I believe that the current surface angle of each polygon in a scene, rather than its location on its UV map, determines how the reflection map is applied. You might think of it as semi-raytracing, but reflecting from a map rather than the actual scene around it. It will change with every pose, so there is no template for reflection mapping. (Or at least this is the illusion that I have been operating under and it seems to produce results as anticipated.) Someone had done a nice illustration of the concept with the object to be reflection mapped - located inside a sphere that displayed the reflection map. Anyone recall where that thread was? I'm not sure a verbal description will work well here.