Except that there really
isn't imprecision in the estimates. The numbers are really quite precise and the confidence interval is actually very good. Far better than an XKCD style should be used for.
I use XKCD stuff for WAG's, not well-supported data.