Despite the fellow being an expert there is no way he can really have flywheel results. It is simple - the only way to have that is to have complete information about entire drive train loss (less the tires since it is a hub dyno) as a function of rpm. The only way to really get that is to test the drive train in isolation or to have the actual at the flywheel dyno results and the hub results and subtract. If he doesn't have that then he is estimating drive train loss.
Of course this contradiction weakens my point above about evidence for under-rating....