Hi Simon,
I've attached two images that I took this past winter of the double cluster M35 and NGC 2158 in Gemini. I used my little 71mm f/5.9 achromat refractor and a 10-year old Atik 314E CCD. M35 consists mainly of bright blue stars and NGC 2158 of faint red stars. It took several nights to capture due to my east-only facing opportunity window and my intended objective of achieving a low noise image. I used Optolong LRGB filters.
I used AstroPixelProcessor (APP) for calibration, registration, stacking, color combine, and final stretch. As with most software the intermediate result is four stacks: one each for LRGB. At that point you can decide if you want to create a full color image or a monochrome image from just a single stack. If you want full color then you use APP's Color Combine tool but if you want a monochrome image of just the luminance channel then you just open the luminance stack file and stretch it.
The monochrome image that I attached is a stretch of my luminance stack. The color image is the result of combining the four stacks and then stretching it.
Notice that my monochrome image is not the same as a gray scaled version of my color image even though it may look that way. There is some extra work that APP performs in order to combine the stacks into color. Ideally you want the signal-to-noise ratio of each of the four stacks to be the same or approximately so. In the case of your NGC 5906 color image I calculated the following stack SNRs:
L: SNR 47
R: SNR 22
G: SNR 24
B: SNR 18
There are two problems:
1. The blue stack is significantly weaker than red and green but this can be corrected with your software's color calibration tools.
2. The biggest problem is that the color channels are much weaker than your luminance.
Regarding that second point, remember that luminance is there to give your color "brilliance". If you've got poor quality color then no amount of luminance can save it.
The root cause of the deficient color channels was the choice of exposure time. From your comments you mentioned 5-minute subs for RGB and 10-minute subs for L. As a general rule I would flip that around so as to take 10-minute RGB and 5-minute L since your color filters are letting in one-third the light of luminance. I take it a step further and set RGB exposure to be three times that of Luminance, so that would mean 15-minute RGB and 5-minute L. That way, in a single sub you've achieved SNR parity across all four channels. The final step is to follow the guidelines that RoboScopes gives with respect to color balance. For example, the KAF-16803 sensor requires you to capture 20% more red frames than green, and 33% more blue frames than green. (This is needed due to the sensor's QE curve.) After that you are now assured that the SNR of each of your four stacks is approximately equal. You will discover that post-processing will become a breeze.
Brian
PS: Please forgive my fat blue stars. If you buy a refractor, make sure it is an APO!