Hi Kevin,
That's a nice image.
Your process seems quite sound. Registering all the subs to the same reference is perfect and using LinearFit to equalise the three masters is good practice (unless you like pure green images.)
Here's my rough workflow for Hubble NB...
Prep: calibrate, register & stack as you're already doing (hopefully you're tweaking the rejection algorithm and parameters to get best SNR from the data?)
Colour data: Remove stars from a copy of the NB masters (star mask & repeated application of MMT to remove first few layers then CloneStamp clean up.) MLT noise reduction (you can go heavy.) SHO colour combine and apply unlinked HT stretch. ACDNR noise reduction and lots of ColorMask/CurvesTransformation tweaking to get colour as wanted.
Luminance: Use ImageIntegration without rejection to do a noise weighted combination of the NB masters. This is the luminance. MLT noise reduction, HT stretch, dynamic range tweaks (HDRMT, LHE), star reduction if needed, more noise reduction if needed, sharpen if needed (clipped lum mask and MLT bias adjustment)
Final image: LRGBCombination to combine Lum and colour. Final tweaking.
Voila!
Cheers,
Rick.