Paul Edwards at StrathClyde, who has added a CL spectrometer to his Sx100 in place of the PMT, writes:
Yes, I use the Cameca optics and the PMT port. However, I don't use a fibre at all, but directly focus the light to the entrance slit of a small (0.125 m) spectrograph which is bolted to that port. I use a simple achromatic doublet lens to focus the light. I don't remember the focal length of the lens I use, but it was chosen to match the f-number of the spectrograph (f/3.7, or NA0.135). You will need one which matches the acceptance cone of your fibre, which is probably NA0.22. If I remember right, the rays are diverging by the time they get to the port, as the image plane lies some distance into the optics assembly.
On thing to bear in mind when choosing a fibre is that the diameter will determine the field of view of the CL. In my case, the image formed at the detector is magnified by around ×3, so when I use a 25µm spectrometer slit I have a field of view of around 8µm. Of course this is not a problem when acquiring point spectra or stage maps. I think in your case the the total magnification should be nearer 2×, as your fibre will have a larger acceptance angle than my spectrograph.
I focus the light to an image at the entrance of the spectrograph, whereas you would focus it to an image at the entrance of the fibre. I just called it "directly coupled" as I don't go via a fibre. In either case you are collecting a lot more light than if you were to just stick the end of the fibre into the port. I avoided a fibre because this would have lost some of the light, but I can see how the flexibility (literally) of a fibre-based system would appeal.
Please keep your software updated for best results!