Although I find it a good approach to get objective measurements of this issue, I would be cautious about using only one measurement tool. Fraps is installing its own function hooks. The hooks are simply JMP operations assembled ad-hoc at the beginning of Direct3D API functions. It may have a small or big influence on the results, especially if you compare with other games rendering with different APi like OpenGL.
Fraps is certainly a good tool but it may be not so suitable for some of the configuration comparisons.
I am not saying that there is no isssue though. Just that it is not so obvious and involving other parameters than Fusion rendering.
Debugging such issues can be really time-consuming and may involve changing third party drivers like rendering dlls.
In any case, I think taht Mathias' approach is worth the try.