With version 7 we now have fully working perspecitive projection orientability maps.
The advantage of using orientability maps compared to using 100's+ of vertices per model is the great reduction in matrix multiplication, along with unlimited detail and other mentioned benefits.
Check the "up_info2.txt" example included in the "renders/" folder for an example of orientability maps in action.
Note: although a matrix multiplication is used at the end of the perspective fragment shader to obtain fragment depth, this can be simplified (eg, using a dot product, a scale and offset), resulting in only four matrix operations in the vertex shader. Parameters such as search jump offsets, minimum discard distance and number of search steps could also be adjusted based on object distance, detailization, and other factors.