I have an homemade flat panel with few blue LEDS and several incandescent normal car bulbs (all 12V), with many covers of paper and acrillic. The amount of light for my QHY 268M CMOS is really small, but for luminance filter it's huge…I did an experiment with lesser exposure times but even at times <1 second it saturate the sensor.
So I tested taking flats with the front of the scope covered (light zero) and discover the sensor still have some signal (!) but the flats were too dark and useless…
At last I go the opposite way and give more exposure times, about 2 seconds for luminance, 4—11 sec for red-green-blue filters and about 30 sec for narrowband filters, as an old flat box I made that works well with my last CCD.
Today I'm using the times commented here, regardless the saturation (high ADUs) of the sensor, and surprise…these work fine !
Pedro