I've done further experimentation with my fibre optic approach and the mirror screw. The individual fibres are visible in the mirror screw, and I'm convinced that some sort of diffusion material is needed to blend the colours. This will significanly decrease the light output, since the diffuser will have to be placed some distance from the ends of the fibres to avoid hot spots. I think that Steve's suggestion of using recessed LEDs will probably result in a higher light output, but a diffuser will be needed at the slot.
I've ordered some high power 5mm LEDs (9000-16000 mcd) which I'll try.
http://www.earlytelevision.org/color...crew.html#1-4a
http://www.earlytelevision.org/color...crew.html#1-4b