Does Levitt have any sampling data or does he just have an idea.
It should be a trivial matter to get a few hundred GMs and IMs to either submit the IQ scores from their school days or take a test and then do a little math to see if the idea holds water or not.
I strongly suspect you wouldn't have to look far to find people with IQs in the 100-120 range who had ratings in the 2600 range.
They are trying to say that 'given many years of intense effort will tend to equal approximately'.
http://www.jlevitt.dircon.co.uk/iq.htm
Or 10 x IQ + 1000 for FIDE rating.
It's far too general really but I guess only meant as an approximation.