My process for this in PixInsight was:
1) Perform all frame calibration (flats, bias, darks) on light frames
2) Debayer all light frames
3) Use batchchannelextractor to extract red files into one folder and green/blue files into another
4) Use subframe selector to analyse the red files and the green/blue files, if blue files contained enough SNRweight and overall signal I'd include them, if not I'd discard them. Output frames with weighting/FITS keyword
5) Use staralign to align red files to the best subframe from the red analysis and then use green reference frame from the same red frame source (so if the red frame came from Image #1, use the green frame from Image #1 to align the green/blue to) and align the green/blue frames.
6) Stack red frames (name the red stack H-alpha) and green/blue frames (name the green/blue stack Oiii)
7) Do all my cropping, background extraction, and stretching/curves on the H-alpha and Oiii, be sure to match the crop.
8) Use PixelMath to combine the H-alpha and Oiii.
For step 5, if you use difference reference frames for the red and green/blue then you'll have to align the stacked H-alpha and Oiii frames using staralign during step 7.