There's certainly evidence that the thresholds of that "accounting" can be modulated by the brain. We know that hypothalmic hormone secretion changes in response to starvation, for example. But [edit: that modulation still acts on, and] still always requires, input from other parts of the body, particularly metabolic signals (though different signals than in e.g. muscle fatigue).
That's where I think we do have differing interpretations of JE's argument. Back to the walking example, we know that the brain can't accurately gauge physical exertion if you block input from metabo-sensing afferents. That is actually contradictory to what JE was arguing, which was that the "tallying" must happen from inputs that aren't metabolic in nature, based on the assumption (I do not share) that peripheral metabolic state isn't changing "all that much"--if I'm understanding correctly