I think you also have to consider what's the more likely.
For example,
Say you have values of 80, 75, 73, 69, 45, 47, 48, and 50. The average is about 61...but in reality there is just of good as a chance of seeing 70+ as there is 50
Anyways...we're dealing with a very large data set so the spread may not be that crazy...but it would be really awesome to dig into and look at the high temperature for a particular day back to records and then do a breakdown.