Some new research on metabolic cost of running/walking up steep gradients: press release here http://www.colorado.edu/news/feature...uphill-running , from where you will find a link to the published research paper (but you will only be able to read the abstract of the paper unless you belong to an institution that subscribes to Journal of Applied Physiology). The experiments were done with runners on inclined treadmills, which obviously has some differences to running up real hills, but the results are consistent with what I found when I analysed race records from uphill (and downhill) races. Essentially, given the choice of going straight up a steep slope or taking a longer route at a gentler gradient, go straight up (at least on slope angles up to 35 degrees, and there aren't any fell/mountain races with sustained slopes steeper than that - are there?).
The paper also includes an interesting discussion of running vs walking. The metabolic cost of walking is less at slopes from 15 degrees upwards (at the fixed vertical velocity used in the experiments), but many of the experimental subjects preferred to run at some steeper slopes. This is certainly a topic needing further research.