For my mosaics, I try to do them both ways and see which looks better. For method 1, i combine them early in the workflow then do the processing steps on the combined image. This will work fairly well if the two were reasonably close in brightness, contrast and noise levels to start with. You may want to add Linear Fit to your mosaic workflow, it will help equalize the background levels.
Some mosaics I find that processing them separately works well and then combining them toward the end. For either method, a bit of tweaking may be required after combining. Once you go non-linear, keeping the brightness / contrast in equilibrium becomes more of a challenge.
