Is it surprising that items in range 64 - 120 is larger than 0 - 5?
We can naively estimate the relative size of these cohorts like this:
>>> math.log(6) - math.log(1)
1.791759469228055
>>> math.log(121) - math.log(65)
0.6214032757011045
So we should expect a little under a 3x difference in favor of 0-5, which is in the rough ballpark of the 2.6x difference that the chart shows for 1950.Note that actual scientists work on this, and use much better methods than this (with confidence intervals and all that jazz). However, this method is a decent "try it at home" approximation, and still much better than comparing the number of years in the age cohorts.