Yikes, my bad.

The acceleration depends on all the forces acting the riders, including gravity, rolling resistance, and aerodynamic drag.

In other words the net force acting down the slope. Not just the force of gravity. Heavier riders generate a higher net force on a similar bike.

free body diagram and all that. college physics was a long time ago.

no

Maybe I'm an outlier, but constant power all feels the same to me*, as long as I can do it at a comfortable cadence.

* with respect to perceived effort.

I have a theory, and I have not studied physics formally, so it could be way off. I suspect the difference is that the force required is closer to constant throughout the circle of the pedal stroke whereas without a headwind, you can apply less force in the dead parts of the stroke. Does that make sense? In other words, there is a sine wave graph of the power input into the pedals, and it's flatter when riding into a headwind. Watts averaged over small intervals will be higher. I can't express this properly but maybe you get the idea.

Why would the forces be less in different parts of the pedal stroke?

I has questions.

Why is the black curve higher average watts but lower % VO2 max?

Does "one leg" mean pedaling with one leg, or pedaling with two legs but only measuring one?

Does a nonzero torque at 200 degrees (20 degrees past BDC) imply pulling up?

That's the force (torque) produced by the rider. The question is why would less force be required during different parts of the pedal stroke, i.e. why would the external forces (rolling resistance, aero drag, etc.) vary with pedal position.

I've slightly re-ordered your question.

Only one leg was measured.

Yes.

There are two possible explanations. First, L/R asymmetry presumably varies with power so although single-leg power wasn't much differernt, total power could have been; but, second, since total combined power of both legs wasn't measured, it could be that pulling up was so much more taxing on this particular rider that his economy was way lower. Or some combo of both.

Good point. There are minor differences in aero drag between having your legs at 3/9 o'clock vs. 6/12 o'clock (which we can spot if you're trying to estimate CdA by coasting) but, absent that, the demand side should be the same.

OTOH, if you're pedaling, the crank inertial load can be different, so you could (probably do) produce different forces around the pedal stroke.

[Edited to add] But to get back to the OP's original question, we know that riding indoors and outdoors feels different, even at the same power. The main differences are differences in crank inertial load and cooling. It seems reasonable to me that X watts in a headwind will feel different than X watts in a tailwind because I've perceived differences indoors and out.

Monentum loss during the low torque part of the pedal cycle. Which then has to be recovered with more pedal torque.

This is most pronounced when doing low cadence up a steep grade. It can feel like the bike stops at the bottom of the pedal stroke.

Okay, but I think the assumption is that the two cases involve the same power and cadence.