the clipping is happening because the na of the collimator is not up to the task, so a prism wouldn't make a difference. I think Simon P1t8ull tested the 250mw 642nm a while back and got pretty poor results.
Printable View
I have a couple of the Mitsubishi ML520G51's here which are the 150mW/640nm small emitter diodes with 7/40 degrees divergence. I tried various FL collimators from 3.6mm to 8mm on the diode with a set of prisms and also found the beam wasn't all that great with any of them. With prisms the losses were unacceptably high due to the severe angles needed to get the beam under any form of control as well, so no good IMHO.