Yes, the results seem to be a little enigmatic.
For a first look-see chose what appeared to be most problematic area - a low fannage curve.
At equilibrium
Radiator heat in = Radiator Heat Out
Wwi + dPwQw = Wa + Wwo
TwiCwQw + dPwQw = Wa + TwoCwQw,
Wa = (Twi-Two)CwQw + dPwQw
The imaginary + or - 0.02c error and/or the inclusion of frictional heat (dPwQw) in the heat dissipated seem to help a liitle..
An 0.02c offset error (represented by top and bottom of error bars), on its own , is no better.