Take note that using your MM( (mean maximal power) data can work okay for efforts like this that are performed at (relatively) the same intensity throughout until failure is reached. However, where you’ll start to see that method failing is with more complex patterns of fatigue, such as stochastic group ride (or race) data - check out this other thread: Xert Breakthrough TP vs Ramp Test discrepancy - #22 by ManofSteele