The heavy lifting seems to be done by the study linked in the "anatomical studies suggest peak speeds up to 15.6-17.9 m/s (35-40 mph) are achievable" line. I'm not sure where those exact numbers were pulled from - I can't find them with a cmd+f. One line in the study uses some nearby numbers:
> If, for simplicity, we assume no change in contact lengths or the minimum aerial times needed to reposition the swing limbs at top speed, the average and greatest individual top speed hopping forces (Favg) of 2.71 and 3.35 Wb would allow top running speeds of 14.0 and 19.3 m/s and of 50 and 69 km/h, respectively
But the study concludes that, even though our leg extensor muscles can produce much higher maximum forces than those generated during sprinting, the "contact length" imposes a constraint on these "hopping forces":
> Because humans have limbs of moderate length and cannot gallop, they lack similar options for prolonging periods of foot-ground force application to attain faster sprinting speeds at existing contact time minimums. Consequently, human running speeds in excess of 50 km/h are likely to be limited to the realms of science fiction and, not inconceivably, gene doping.
So the craziness of the original estimate seems to follow from on a misreading of that study.