Is a 6% rejection rate when stacking flats still acceptable?

Hi there, I am following the manual preprocessing tutorial for my first time working with calibration shots and got a bit confused when I got a 1 - 2% rejection rate while stacking the biases. A quick google search revealed that the 0.1 - 0.5% rejection rate mentioned in the tutorial seems to be outdated, I could not find any suggestions on what rejection rate seems acceptable when stacking calibration shots.

Now I’m encountering a 6% rejection rate when stacking flats:

17:48:23: Starting stacking...
17:48:26: Pixel rejection in channel #0: 1.935% - 1.506%
17:48:26: Pixel rejection in channel #1: 3.191% - 2.838%
17:48:26: Pixel rejection in channel #2: 23.675% - 6.345%
17:48:26: Rejection stacking complete. 128 images have been stacked.
17:48:26: Integration of 128 images on 128 of the sequence:
17:48:26: Pixel combination ......... average
17:48:26: Input normalization ....... multiplicative
17:48:26: Output normalization ...... disabled
17:48:26: Pixel rejection ........... winsorized sigma clipping
17:48:26: Rejection parameters ...... low=3.000 high=3.000
17:48:26: Creating rejection maps ... no
17:48:26: Image weighting ........... disabled
17:48:26: RGB equalization .......... disabled
17:48:26: Background noise value (channel: #0): 20.363 (3.107e-004)
17:48:26: Background noise value (channel: #1): 21.678 (3.308e-004)
17:48:26: Background noise value (channel: #2): 21.848 (3.334e-004)
17:48:26: Saving FITS: file pp_2024_07_06 Saturn_2 flats_stacked.fit, 3 layer(s), 1920x1080 pixels, 32 bits
17:48:26: Execution time: 5.12 s

Is this still acceptable?

Further notes:
I am shooting with an astromodified DSLR (EOS 77D) that has about 10.000 shots. 128 Shots where each taken for flats, darks and biases. I have stacked the biases using Average stacking with rejection, no normalisation, winsorized sigma clipping (3.0 for both low and high sigma). Flats where calibrated using the master_bias before stacking - short: I followed the tutorial as best as I could…

Images were taken using the 77D’s time lapse function, which provides an .mov file instead of a pure RAW-image - could this be the culprit? (ISO 1600, 1/50 exp, 1 shot / second).

Hello and welcome.

How do we can now? The best thing to do, IMO, is to output a rejection map to see what it is rejected.
Also 128 subs for flat is way too much. These frames are taken under good condition of light. There is no noise. Only a few is needed.

It is not good at all. As I see, you have already debayerd images, where we want RAW and not debayerd images for calibration step.

Hi and thanks for the fast answer. I will retake my calibration shots as RAW.