The concept of using reaction torque for a dyno has LOTS of advantages
and, if properly executed, works well and is robust and safe and stays
in nice calibration over extended periods of time.

However: The effects of the prop wash on the airframe will be so profound that
the dyno "results" just due to the aerodynamic errors will be no better
than +- 15% - - assuming every other error in an exhaustive error budget
is zero.

Ask me how I know ?

George Braly (PS. Just spent two hours in the test stand, running with reaction
torque from a propeller doing detonation sensor calibrations - - mode
<g>.)