I thought that number was a much stricture control. I think you have to do a lot more go get to the number you're talking about and there's no explanation for that one. It's been a bit since I looked but I don't think things improved that much. Even still, a 1% median difference on earnings is a pretty large number.
Since we're talking about a median, it would be interesting to see how many people you would need to remove from the top end to get the numbers to line up.
Compound that over a person’s working years and it adds up to a non-trivial number at the individual level. At the aggregate level it’s already non-trivial without any compounding or the like.