Well, I don't know (you guessed that :o) ). I'm just thinking of kangaroos whose achilles tendons recover 95 % of the kinetic energy so that, only slightly exaggerating, to get from A to B at a constant speed they only need power twice, once to start, the other time to stop, regardless of how long the distance actually is. If that isn't just another urban myth, of course. :o)
When I mentioned cartilage, I assumed it was elastic enough to help make the leg jump up again when it was rammed against the ground, and might therefore inflate the error bars. No moment involved, I think. But as cartilage is often described as shock-absorbing, that probably means it is just not elastic (enough to have an effect here). I don't have any numbers at hand. Can someone help?
The idea suggested by HP Mike Taylor sounds very plausible. When I run for a long time and am exhausted (such when I've missed the last possible bus or when I had to run 700 -- 2100 meters at school), I regularly do this intermittently (means, I just lean forwards). That way I get faster with the same effort. But it doesn't work the other way around, less effort at the same speed, because there's a stalling speed. This is also the reason why I can't sustain this when I'm exhausted (haven't tried yet otherwise, now I think I should): I get faster and faster and can't really keep up, so I have to stop it after less than a second.
The idea is not entirely new; the big paper on Pristichampsus rollinatii suggests that at high speeds this croc...odyliform :-) was able to run bipedally in spite of its center of gravity lying in front of the hips because of this effect, and the paper that suggests Giganotosaurus was able to reach about 50 km/h (see www.cmnh.org/fun/dinosaur-archive/2001Oct/msg00107.html) seems to rely on it, too.