I'm less interested in the worthless calculator and more interested in the paper's argument that relative muscle mass in legs is a key determinant (more mass = more glycogen storage). Marathon running takes very little power (hence not much muscle) and increased mass on an accelerating limb takes more energy. The distribution of the mass is important; way up in the quads/hamstrings is better than in the calves. Anyway, the model is pretty simple, has some interesting parameters but fails to account for (at least I didn't see this at all) biomechanical or muscular efficiency.