Just sent a 918 patch target though CS5 and CS4 identically (don’t have CS3 installed, I’ll do it if you insist). Used ABW, measured on the iSis, Avg dE is 0.68! Let the target dry down, that will go a lot lower (these prints are fresh out of the 3880).
So No, again, terms of your post, there is nothing broken in ABW in 10.6 in CS5. But there could be something broken in your beloved QTR profiles or how you built em, used em etc.
Use ABW correctly as designed, and works as it should. In CS4, and in CS5. No reason to believe CS3 is any different and the original post clearly put the blame on CS5.