I sure this has been covered before...but worth asking again. Perhaps Invigorator has changed.
I am trying to do a reflection map of a single word on a rotating cube. Problem seems to be the control of how the image is projected. Currently there seems to be no way for me to map the word cleanly at a size that works for me. The map envelopes the entire model and is extremely large.
You might try making a precomp in the same aspect ratio as the cube face and placing the text in that precomp. Changing the size and position of the text while leaving the dimensions of the precomp alone should give you limited control of how the reflection map appears.