The numbers are based on testing on visitor's computers, and the IE number in particular is based on a sample size of 1 (at the time that I'm writing this). So maybe the guy who was using IE was doing so from a very fast computer? That said, I got over 50k ops/sec on my laptop using Firefox and less than 10k using Chrome.
I ran IE and Chrome on the same computer. In the end I went with a function that has the loop inside it and it's fast enough. I'm just surprised it's so slow.