there are so many ways to skin this cat... the most straightforward is what you describe first.
there are lots of variants on the 2nd/3rd option. one would be to use superpixel debayering on all the subexposures (bias, dark, flat, light), then extract the red channel from each of them and throw away the others. now you've got files that are as though they came from a mono camera. so you just put all those single-channel files into BPP and go from there.
if you don't want to use superpixel debayering, then you could use BPP to calibrate the raw images, then run ChannelExtraction on the calibrated lights, then put those single-channel files into BPP.
the advantage is mostly just that you will save a lot of disk space and the runtime for the whole mess should be shorter, as PI only deals with 1/3 of the data. but the results in all 3 cases will probably be the same.
rob