This is the idea of progressive overload that has been a core tenant of many weight training programs for a long time. It may be that body composition changes are primarily adaptive responses to stressers, rather than the typical simplified model of 'x work burns x calories'. That is, x work may in some cases be associated, on average, with x calories being shed, but it does not directly cause that outcome in a straightforward fashion. Rather, it may well be the case that your body adapts in proportion to the level of stress that it experiences (that would be consistent with the conclusions here).
Another consideration is that when we assume that natural processes respond linearly, we seem to have a tremendous track record of being wrong. Even with that being the case, essentially every model that covers any aspect of fitness seems to assume linear response. So it's not surprising to see demonstrations that these models don't hold to reality at the extremes.
Since the Fitbit was released, I've maintained that counting steps is a useless activity unless you are very old or infirm. I think that practitioners (i.e. athletes) have long known that intensity is the primary determinant of changes in body composition and athletic performance. It's seemed to me for a while that one of many causes of the difficulty people have improving their bodies is that they 'exercise' rather than 'train'. That is, they think of their workout as putting in time or quantity of work, rather than just trying to improve their performance across various physical tests.