Output Normalization

I am stacking images with Siril (both 1.2.0 and 1.2.1) for photometry but the output never reflects the full range of the input subframes. This voids the EGAIN setting in the FITS header which is often needed for calculating S/N. (Although the stacked image still includes the now incorrect EGAIN from the subframes.) This happens regardless of what stacking method or input normalization I use (I generally use average stacking with rejection, normalization additive with scaling, Winsorized Sigma Clipping, no image weighting and output normalization button unchecked.) The state of the output normalization button is correctly noted (well, it actually says ‘nomalized’ or ‘unnomalized’) in the FITS header by Siril even though it makes no difference to the total range of values in the image. (However, the output normalization button does something like equalizing the histogram - darkest pixels are still zero, brightest are still 1 but mid-range pixels are greatly brightened. Like pulling the grey slider way up in Photoshop.
Am I missing something? I seem to remember once seeing something in the docs about some detail in the image processing which enforces output normalization but I can’t find the reference any more.
Images are from a QHY183 mono camera, gain=0, offset=5, unbinned, acquired using NINA. This is a 12-bit camera with ADUs scaled up by 16x to give a 0…65536 range.
Otherwise I love the program for my occasional pretty pictures.

Welcome on the forum,
I’m not sure what you are trying to do, photometry is usually done on unit images, not stacked.
Anyway, when reading an image with the EGAIN keyword, Siril will store it in next images as CVF, it’s not lost.
For stacking output normalization, documentation states:

If Output Normalisation is checked, the final image will be normalized in the [0, 1] range if you work in 32-bit format precision, or in [0, 65535] otherwise.

I sometimes shoot extremely faint targets and cannot do quality (or any sometimes) photometry on a single image. Or I am shooting targets which are very bright so I have to stack multiple very short exposures to reduce scintillation noise. All my reduced images are in 32-bit mode so there is no option there either.
I saw the EGAIN is retained as CVF. Except that once the image is normalized that value is no longer applicable.
Either way - turning on or off the output renormalization button does not prevent the renormalization.