Power 3 is the best. I don't really see much point looking into that further, but you can if you want! They are different algorithms going from basic and simple processing to more advanced and complex which is more effectively tuned to the human ear. I have not read up about it for a while so a bit rusty. Someone will give better info.
Dither is the VERY LAST PROCESS you apply before turning into mp3 or burning to cd. You apply when you do the final conversion from 24 to 16. Often we do the sample rate conversion at the same time and that should be fine. SONAR should do that first, then do the bit depth conversion after, but in one mouse click.
I have done experiments with it before. The dither floor is so darn low that I see it really making only a teeny difference. I don't think it's anything to stress about at all. Spending more time to fix a hiss issue in your recording chain will yield far bigger quality changes.