Dual band drizzle

Whenever I use a dual band filter with x-trans drizzle and extract Ha and O3, I'd really like to understand why I see much much worse drizzle artefacts from the O3. With my setup (undersampled and well dithered) after about 100 frames I'm starting to get good results from the Ha with no noticeable remaining artefacts, but on O3 I'm still having problems with 2x and 3x as many frames. I know I can fix this with more data, but why should I need so much more data to get good O3 drizzle results when Ha is fine with less?

The exposure time is the same for both, so I must be missing something about the x-trans algorithm of how Ha and O3 are extracted which is meaning there's much less coverage of the grid for O3? My understanding is that just because O3 is weak in the target this shouldn't make a difference - the artefacts in the drizzle aren't about how strong the data is, but about how much coverage of the array there is? If it was about the strength then black sky would always show lots of drizzle artefacts.

If I was extracting RGB I'd expect green to show better quality than the others because of the doubling in the CFA, if I understood the O3 extraction algorithm better would this explain the problem with O3? On the face of it Ha should be worse because it's mostly the red channel, whereas O3 is drawing on the other two?

Or is there a bug in APP here I'm experiencing? I'm really puzzled by this behaviour and would really like to understand why I can get great Ha extraction but terrible O3 for the same exposure time. Thanks.

-Mark