CG also plays a part, as a rearward CG reduces rudder authority.
Correct, but as far as I know, VMCA, VMCG is determined at the most unfavourable CG location (Aft limit)
I think that JAR state that in the determination of Vmca and Vmcg you have to consider that this are Calibrated Airspeeds
CAS is IAS corrected for instrument error (internal inaccuracy) and position error (measurement of static pressure) so as said before density plays no part here.