Actually an average of high and low, statistically speaking, is an statistical measure of the mean, but a much more inexact one. Did that problem once in a stat class, and the variance is huge, the sqare root of two versus the square root of say 24. That's what 16 times bigger. But that's only half the issue. What's the higher order stats--skewness, kurtosis. Plus who knows much about the underlying distribution. Here, in Iowa, noteworthy for changeable weather, we can get a morning very warm, and an afternoon from another distribution, a front goes through. So the high can be from Mars; the low from Venus lets say. Doesn't even make sense to average them then. And tree rings don't even show high or low, and not the winter.