I suppose in a perfect world, maybe one uses proprietary pads for each rim, and designs the system to use constant input watts but modulate brake caliper pressure to achieve a constant wheel speed, and that would be an apples-to-apples comparison, i.e. it would tell me how rim A (paired with brake pads A) would compare to rim B (paired with brake pads B) at getting me safely down mountain X. But that would require a way more sophisticated test setup that could adjust brake caliper pressure to achieve a desired speed.

I suppose a cruder version could involve manually calibrating the brake caliper pressure to achieve a set wheel speed at the beginning of the test, and then letting things run. As the brake pad and rim were progressively affected by wear, wheel speed would start to deviate, but this would at least put the rims in the ballpark of each other, so it would be more of an apples-to-pears comparison, better than apples-to-oranges.