The test is whether you generate shares faster. The MHash/sec number displayed is just eye candy. I can change the code to display THash/sec instead if I want. But if you can convince the pool that you generated more shares, then you have something. Time how long it takes to generate a fixed number of shares (at least 10) with and without your change.
Yeah, that's how I figured out that it didn't really generate any more accepted shares than the nonmodified miner. Well, I hope I'll learn something from this - never go public premature.
Again, sorry for all the commotion.