Debayering and master dark scaling

Hi,

About 2 weeks ago I posted

    Problem calibrating with demosaic narrow band data

and within about 5 hours Mabula had explained my problem and how to fix it. So I can't fault the service I get here!

Now that's great, but the more I thought about it, the more I realised that there were aspects of the problem that I didn't understand so I have a supplementary question. It may be academic but part of the fun of our hobby is understanding the technology which is in any case key to best exploiting it.

I will briefly restate the original problem, albeit with different data

My setup
Target M42 170x15sec 125gain
Capture RedCat 51. Optolong L-eNhance Dual Narrowband. ZWO ASI183MC Pro
Processing MacOS Mojave. APP 1.083 or 2.0.0-beta12
Calibration Full set BPM,MF,MD(300sec),MDF,MB

My problem is that when I debayered with the algorithm Ha-OIII color my integration looks like this (maximally overstretched to enhance the effect).

Clearly horrible amp glow even though I thought my calibration was fine.

Mabula explained that amp glow isn't linear in time, which does make sense, so dark scaling won't work; I need to make my MDs with 15sec darks.

So, I have a solution, BUT if I integrate with the Adaptive Airy Disk then I get this.

which looks like the amp glow has gone! I appreciate that comparing stretched histograms can be misleading so I saved my images un-stretched and then in photoshop carefully stretched both in exactly the same way and the result was the same.

Naively, I thought that as dark noise isn't "coloured", it has not been subjected to any optical filter, its subtraction ought be more or less the same regardless of the debayering method. Clearly I am mistaken, but I should really like to know how I appear to get away with scaling with AAD but not when I use the correct Ha-OIII color?