There is no simple way to do what you want. The Virtual Power curve for your trainer is set based on testing in TR. Differences can exist between the tire and wheel setups that may yield different actual power values as the given speed.
Because the tire in particular can influence the testing, for best chances of getting good data, you should install the power meter on your trainer bike for your testing. This is helpful to reduce the number of variables between the power meter and the VP value from the speed and trainer.
There is a way to manipulate the results, but it is a hack, and will likely still not “solve” your problem.
In short, you can adjust the wheel circumference setting in TR to alter the Virtual Power value. You can increase or decrease it and it will effectively change the value.
As with any two power devices, they may have a different slope. This means is is possible to have them align at one power value, only to have them diverge and have different values above or below that point. I suspect that is VERY likely in this example, but it’s all speculation.
With your trainer bike and power meter temporarily installed on it, connect 2 devices.
- Power meter to your head unit.
- TR on your device with Virtual power.
Then ride and see what power reports for both devices are various speeds. I would set power smoothing on TR to 3-5 seconds and head unit to 5-10 seconds (since it will be more jumpy).
Hold VERY steady power and cadence at a range of power levels, below, at and above your FTP. Compare the differences as a percentage and see if it is reasonably consistent.
If they are, great. If they aren’t, that is the mismatched slope issue, and you may want o focus on the area around tour FTP.
Once you have a percentage difference. You can adjust wheel size in TR, and then retest pedaling at the same power at your target pedals. Evaluate if the wheel size change moved the Virtual Power in the right direction.
It really is a hack and will not be perfect by any means, but it’s how I think you can get something better.