Something odd with the corners of subs from Pier 5

So this is weird.

Some members of Stargazers Lounge pointed out that in my NGC2237 (Rosette Nebula image), the corner stars were weird, and suggested I was using BlurXterminator too aggressively. Rosette using HaO3 and S2O3 filters on Pier 5 (roboscopes.com)

The top right hand corner of the calibrated and stacked image from WBPP shows little bumps on the stars at about 11 o'clock. They do not seem to be classic egg-shaped stars, but smaller 'second centres'. Applying BX resolves them into two stars, a bright main star and a much smaller/dimmer second ghost star.

I though that this was some sort of registration error in WBPP, but looking at a single sub, the same effect is there. So it must be either optical or in the electronic readout of the CMOS chip.

This is with the HaO3 filter (same effect pretty much with the S2O3 filter).

Could it be some internal reflection in the dual band filter, or some interaction with the Bayer filter on the camera chip?

Anyone seen anything like this?

I will look at some datasets from other runs, and also do a channel separation to see if it only affects one channel.