Varying the ADU should never make any difference to anything (aside from noise levels in the flats depending on sub number). It seems to be one of those red herrings that people throw out when they dont have any real ideas, in much the same way as polar alignment gets blamed for everything.
In a perfect world, you are absolutely correct. However, the full answer is that ADU doesn't matter if the flat is exactly calibrated. It will not always be a big difference but is a difference.
A flat frame consists of two main parts -- the illumination signal from the flat source and an offset made up of Bias signal and Dark signal. You could write this as FLAT + OFFSET. If the OFFSET is zero, then when you calibrate, there are no problems and the calibration will work regardless of the ADU level of the FLAT portion.
In that ideal case where the FLAT has had the OFFSET removed in calibration, then when you calibrate your lights you would be doing:
CAL_LIGHT = LIGHT / (FLAT/mean(FLAT))
If your calibration of the flat frame is off, then that calibration factor of (FLAT/mean(FLAT)) will also be off. This can be seen by trying a mathematical experiment.
- Let's say your mean raw flat ADU is 15,000 and the Bias plus Dark for the flat has a mean of 2,000.
- Let's also assume your flats include +/- 15% vignetting from the mean.
- As a final assumption, let's say your light frame has a (dark calibrated) mean background level of 3,000 ADU.
- When you calibrate the flat with your Flat-Dark or Bias, you end up with a mean flat at 13,000 ADU.
- In the ideal case, the Light Frame calibration factor for the flat at a darker point where the actual value of the raw flat shows 12,750-2,000 ADU will be (10,750 / 13,000) or 0.827x. This will cause the image background to be lightened to 3,000 / 0.827 = 3628 to match the vignetting.
Now, let's look at a case where your calibration of the flat was off by 500 ADU due to mismatched dark or other problems. We'll say the calibration took too much of the offset out of the flat.
- In this case, the Light Frame calibration factor for the flat at the same darker point where the actual value of the raw flat shows 12,750-2,500 ADU will be (10,250 / 12,500) or 0.820x. This will cause the image background to be lightened to 3,000 / 0.820 = 3659.
We can see that by removing a too large a portion of the offset in the flat when we calibrated caused the flat to over-correct and lighten the Light Frame a little too much (by almost 1% or ~30 ADU). A similar error can cause under-correction in our images when we don't remove all of the Bias-Dark offset.
You can run through other examples and see how things are affected by incomplete (or over-) removal of the offset in the flat and how that will affect you images. If your mean raw flat ADU level is 32,000 instead of 15,000, you can run through the example above and see how a similar calibration error of the flat will affect its application to your lights. I had written it up as a simple spreadsheet a while back to get it clear in my own mind. Doing so will allow you to just plug in numbers and see how much your lights will be affected. As mentioned, it is a small number but when you get to post processing and severely stretch the image, you will usually be able to see the difference.
You can often recover a poorly calibrated flat that is showing under or over-correction by experimentally adding or subtracting a fixed offset from it and then reapplying to a sample light. Checking the light for over or under-correction will then allow you to adjust the manual offset adjustment amount. Once you get something that looks good, you can then use that adjusted offset flat to calibrate all your lights.
The net of all this is that varying the mean ADU level of the flat will not affect the correction so long as calibration of the flat is accurate. If the calibration of the flat is off due some reason, then the mean ADU level will have an effect on the calibration result. For a given calibration error of the flat, higher mean ADU levels cause less error that low ADU levels in the flat. This is yet another reason to shoot for a mid-range ADU level in a flat so long as there is not clipping or saturation going on.