PI issues with red star blowouts on LRGB combination

[mcore]

I'm having some difficulty from time to time in LRGB combination.  I sometimes get very saturated highlights in my stars, usually at the periphery of their cores.  I've put in an example below - full image for context, then two zoomed-in areas showing L, RGB, and then LRGB.

At first I thought "oh, makes sense, there's an area of intense L that isn't quite aligned with the RGB, so it happens to land over an area of intense R instead of a nice balanced color, and off you go".  Looking at this data though it just doesn't seem right.  The area that turns into highly saturated red doesn't seem any more or less red than any of its neighbors, which didn't get this red blow-out.

I'm posting here in hopes someone says "oh yeah, that's X, here's your mistake."  Any tips?

Small snippet of stars near the center of the image:


Tight zoom on the lower central region:


Full image for context:


Thanks in advance...

[pfile]

it does seem that the registration between the RGB and the L is not perfect, but maybe it's just the gif?

rob

[mcore]

It's odd, it's like the poor registration is only on a few select stars - and those are the ones with red side-blowout.  I'm hoping someone who understands color spaces can give a quick description of why those pixels get red-saturated.  Some more interesting detail for one of the red-saturated pixels:

R:0.0290 G:0.0077 B:0.0042 - RGB value
L:0.0031   - Lum value
R:0.0212 G:0.0000 B:0.0000  - Post-LRGB channel combination RGB value

Why would the G and B channels be totally dropped by the channel combination?  They are 26% and 14% of red's intensity, respectively.  If the ratio held then I would be fine.  Part of me wonders whether I'm looking at some "float" instead of "double" precision issue here...