Again, I am not in the realm of being an expert on this topic, but you can’t determine if the power meter is 100% accurate unless you actually tested the entire system in an experimental setup with a known input wattage/force and saw what kind of response you get. Not to say the Neo or the Favero’s are right or wrong, but it is extremely difficult to say which is which in this case where you have 2 “relatively vetted” data sets that seem off. Note that even crank based power meters have issues, as seen in things such as the new Shimano cranks. Depending on where you place the strain gauge, it will have different characteristics due to material properties (shape, asymmetry etc.)
Again, if you know the offset, just input the offset and be done with it. Again, data accuracy should be very good. Only caveat is I have heard the Neo does have some lower readings when at high flywheel speeds, as there is no strain gauge in the unit, so if you are using ERG its best to use an easy gear and let the device determine the power for you.
And lastly something to consider, manufactures errors may come into play as well. Lets say both power meter manufacturers say they have a 1% error range. I am not sure if this is an error with respect to your “absolute wattage” or your “wattage readings” (ie readings are +/- 1% vs you are 1 watt high or low for an absolute 100 watt value, if both it would be reading 101 x 1.01 or 101 x .99 or 99 x 1.01 or 99 x.99 watts). Drivetrain losses are non-negligible (note quick google :Mechanical Resistance on Bikes: Drivetrain Efficiency & Hub Bearings – Ride Far) so lets estimate 5-7% drive train losses (and worst case you cross chain so lets just say its 7.5%) so now your readings are 7.5% off empirically, even before you consider the device manufacturer acceptable errors. Note this is a napkin calculation.
So now your 100 watts absolute at the pedal will read 92.5 watts absolute at the hub. Note that with errors lets say the pedals read high, will read 101 x 1.01 watts while the neo will read low: (92.5-92.5*.01) x 99 watts (worst case, note we assume the offset error is reading 1% off the absolute value and factoring in a 99% accuracy of the meter, both within tolerance) and we now have your pedal reading 102 watts and your hub reading 90.659 (90.7 lets say). Whats the difference between the 2?
I get 11%. 11% difference. All within known qualities about the 2 power meters. I have no idea if the power meter manufacturers and/or smart trainer manufacturers have some secret sauce to lower the differences, but I would be suspect of that. Do you wax your chain? How good do you clean it? What brand of chain do you use? Do you use high quality bearings? Is your chain sized correctly? Are you cross chaining? etc.etc.
Someone can prove me wrong and explain to me how or why my write up may or may not be the case. If anything this emphasizes how you should use one power source, or understand that offset is offset. Of course this is worst case, but it falls within the manufacturers specs and standards. Thats not to say either power meter or device has anything going wrong, seems like both may be perfectly fine and you are just seeing science being science. If they both read accurately with a measured offset, they are both functioning absolutely as expected.
Edit: It’s worth noting that I believe power meter manufactures and smart trainers are typically 1.5-2% error range, so this will magnify the difference you would read between the 2 meters. End of the day, as long as both power source are accurately recording and the only issue is the offset, there is nothing wrong with the power meters. Throw an offset value and you are done.