They certainly need more data, and could well have conducted the study for hours rather than 5 minutes and might have reached statistical significance (also switching bikes often so that there is no order & fatigue bias).

However, I'm doubting their study is giving any meaningful data for general rideability.

For a track setting (smooth concrete or wood), I would expect a benefit from the stiffest frame possible).
Smooth pavement with the periodic interruption... perhaps also tending towards a stiffer frame.
Rough pavement (chipseal) is annoying, and I could imagine a significant power loss for noncompliant tires/frames.

My guess that their study was like riding in the gutter, or on a sidewalk with a very smooth surface with a few thumps. It just doesn't seem like it would capture the effect of riding on chipseal, gravel, or cobbles. And the smooth nature of the treadmill may tend towards benefiting the stiff frames.