If the differences can't be detected between a 54cm CR1 and a 62cm steel bike, then I think I've already proven my point

I think a better place to cut would be the seat tube and down tube. Cutting the chainstays will allow compression and expansion very nearly parallel to the chain, which would pretty obviously return to the drivetrain. We need to test the kind of flex that exists in real cycling.

I don't see what variables we have to worry about between two bikes. The chain length differences should be minor. We can run the tests with the same gear ratios. We could go single speed with it to eliminate the derailleur pulley differences.

While using real bikes may add a few variables, I think it eliminates anyone claming that a cut frame can't make up the differences with intentional engineering in stiffness improvements. Like it just can't be simulated accurately.

That, and we'd have to buy a sacrificial frame. My Bridgestone is worth as much on ebay as a PowerTap, so it's not getting cut.