I think you could have a good go. You need 2 test frames with different lateral stiffness, a pair of power meter pedals and a rear wheel with a hub power meter. You might as well throw in some crank power meters too for good measure. Calibration is not that critical as long as they are consistent, which good quality meters are.
Then you just ride the first bike through a set power cycle, swap the meters over and repeat on the other bike. Swap back and forth a few times for repeatability. Then it's just a matter of comparing pedal/crank vs hub power loss between the 2 bikes. If there is any power worth actually saving it should show up in such a test.

Someone mentioned confirmation bias, but that would be irrelevant in the above test as you are not asking anything subjective of the rider. All they have to do is ride to the specified power targets.