I think normalize is a good thing, but not always.
Let's say you have a peak due to a crack that you can't recover with Audition. The result would be a "false" normalize between fronts and rears.
It's best that you apply an amplify function differently to front and back channels.
Look in the graphic of your samples, see at what value is the peak (i.e. -6 dB), then apply the amplification to reach -1 dB.
For the rears, it usually sounds quiter than the front, so you need to add 3dB to the values used for the front channels.
3db is the minimum value your ears could notice on a change of volume in sound. If you apply less than 3dB you couldn't notice nothing changing at all.
Another thing, for SQ decoding, it was said initially that the front channels (once decoded) should have been amplified 3dB more than the rear channels, so maybe you don't need this trick if you think that the balance is good enough.
Instead, in Qs decoding script what you get is the volume levels that were REALLY intended to be hear, but if you need to gain fronts or rears it s your choice
Bye