Canon cameras play a gain with bias, I dont remember the exact details, but Craig Stark wrote quite an interesting article on in some years back. I found a link to one of the articles here:-
www.stark-labs.com/craig/resources/Artic...s/CanonLinearity.pdf
It's been some years since i've fussed with a dslr, but from what I remember, there is an easier way to get proper flats using one. Instead of trying to manually adjust exposure times, put the camera into automatic, and let the camera figure it out on the fly itself.
As for twilight sky flats with a dslr, I'm not sure that'll work very well. I played with automatic flats some years back, and came to the conclusion that there isn't enough time during twilight to get flats from all 7 filters, as the light level is changing rapidly, and the narrow band filters take significant exposure times to reach the required levels. But I did find I could 'almost' get a full set if I started when it was still fairly bright out, and did all the narrow band filters first, then the colors, and do luminance last. But I was never happy with the results when trying to automate sky flats. When we were setting up to do a precision photometry run, I could get a really good set of sky flats, if I only did one or two filters, ie the ones we would be using for that nights run.