Nice work Luke.
I learnt a trick the other day (using PI).
Invert, SCNR (remove green), re invert.
Gets rid of the magenta stars.
Or the hard way ( I developed this procedure, well maybe...)
Gather RGB data using short exposures (30 X 30s)
Register the RGB to the SHO
Apply Starnet to the SHO and remove stars.
Apply Starnet to the RGB (registered) and keep the star mask.
Combine both with Pixelmath.
Works a treat (almost, a few stars might need to be fixed).
Have fun