Regression to the mean is ruled out by comparing against strong contemporaries, whose number of papers produced kept growing on average. This group of strong contemporaries was not selected by looking at the number of papers, but instead by looking at the pool of strong mathematicians selected for other prizes.
You are right, but it is not ruled out entirely because they used Wolf and Abel prize in the contenders group, which are awarded for a lifetime of work in mathematics and therefore biasing the contender groups.
And, as I already pointed out, the non-award-winners must by definition be different from award-winners and cannot control for unobserved baseline differences. Even more so if those prizes are, as is only sane, intended to reward different things than what the most prestigious prize in the field already rewards...
