The wall is at 76% of VO₂max, and eating moves it
A marathon is a constrained optimisation. The constraint is grams of carbohydrate, and it decides the fastest pace a runner can hold before arithmetic rather than willpower ends the race.
The arithmetic
Running costs about 1 kcal per kilogram per kilometre. The share of that coming from carbohydrate rises with intensity, so the faster the pace the more carbohydrate each kilometre costs. Muscle glycogen, liver glycogen and what is eaten on the way are the supply. Where need crosses supply is the wall.
What it does not say
It gives a ceiling, not a prediction. Plenty of runners slow for reasons that have nothing to do with glycogen: heat, pacing errors, a lack of specific endurance. The model says only that above this pace the fuel runs out, whatever else does.
Rapoport's formulation with literature constants for glycogen density and the carbohydrate share of energy by intensity, and Keller's two-equation runner for the pacing profile.
Constants fitted to a population do not transfer to a person. Gut tolerance for 60 g an hour is itself trained, and not everyone has it.