Off the top of my head over breakfast…
I suspect the easy answer is that power meter companies have never tested their product with pedal extenders. They’ve likely validated (and internally certified) their products against their own direct-mount equipment and (I hope) a range of commonly used cranks.
Adding an insert of x length, of y stiffness, and installed with z torque… that is likely to changes the force across the spindle/gauges is a step too far for them to officially support.
On a similar/related topic - I’ve recently been using a set of Look carbon cranks with a tri-lobe insert that allows multiple crank lengths depending on the position of the crank insert the pedal screws into. The power numbers I was seeing on the pedals were lower than expected… only by a few watts but when switching out the cranks for a set of Shimano alloy, I was seeing better numbers from the pedals. It was only a few watts… but… More data needed to be confident it was the tri-lobe (it could have also been the cranks which had a lot more lateral flex than the Shimano cranks).