that's the way to go if they have different "peak" values. IOW, if the loudest point in R is louder than the loudest point in L, then L will get a louder volume boost throughout.
What always f's me up is when there's one hit (like an applause) that throws the proportion off.
Also, check RMS normalization. It squashed dynamics, but it might work. Finally, Leegeddy wrote something awesome in a thread once about normalization, and I am quite sure it was saved in teh archive. I didn't understand normalization (and might still not...) until I read that.