My initial claim, which you called wrong, was "The 64-bit product will be faster (like 20%) than the 32-bit".
This claim is true - a product compiled for 64-bit will usually be significantly faster than a product compiled for 32-bit (I've seen quite a few benchmarks, in a range of a few percent faster to over 50% faster - 20% is a "typical" improvement).
The fact that it's due to the changes in the x64 ISA rather than the longer pointers doesn't negate the fact that 64-bit apps will usually be significantly faster than 32-bit apps on the same system. That's all I said, and I stand by it.
I also stand by the opinion that in the long term Apple (and Apple buyers) will regret the "9 months of Yonah". If Apple had waited for Merom, and had embraced true 64-bit for OSx64 (that is, there simply would be no 32-bit s/w whatsoever for OSX on Intel) - things would be simpler and cheaper for developers and users. As it is, there will be another big software transition - from the current 32-bit Intel to a 64-bit Intel, and "fat binaries" will have to become "even fatter binaries" to hold both 32-bit and 64-bit Intel code in addition to 32-bit PPC code.