The
purple fringing Wikipedia article has some good links at the bottom. Not properly correctable digitally (although you can do some gross corrections, you cannot recover the image properly).
The question is how to detect the difference between a) blur due to focal plane being different for the different wavelengths, and b) projected image size different for different wavelengths? A photo that is as in-focus as possible with narrow areas of white (not blown out) against black could help diagnose what was happening - you should be able to recognise what is causing the problem. (a) would be impossible to correct, while (b) could be somewhat corrected.
Maybe try illuminating with different lights? Different coloured lasers would be perfect (not a focused point laser but scene illuminating - trouble is getting hold of a blue laser), LEDs would be OK (they are not a single wavelength so not ideal), or perhaps use broad spectrum light, and use gels to filter everything but a narrow wavelength (I suspect too expensive to get gels of sufficiently narrow wavelength range).
Narrow R,G,B lens filters might work by taking photos the same scene more than once with different filters, but it would be difficult to not to touch anything: the smallest movement would prevent images registering properly (making it difficult to work out underlying cause).
Things I would look at:
1) In your favorite graphics tool, just look at the blue channel and see if it is in focus compared with other channels.
2) UV filter probably won't help (I would expect camera already has UV filtering internally).
3) avoid blown channels (CHDK helps here with zebra mode configured to show blown colours).
