I use curves and levels on the L channel of the L*a*b model as a standard procedure. Curves used in RGB always produce a colour shift, this can be more or less pronounced, but it's always there to my eyes.
I also find ColorPerfect by CFSystems a very effective piece of software, even though with the most contorted user interface and user manual I have ever come across.
One day I'll have my proper darkroom and I will be able to judge results by optical printing. At the moment I cannot judge colours in negatives on their own, it's only the scan I see. Sometimes life is much easier with slides.