On the usual IQ scale 145 is 99.865 percentile. In 2000 (50 per year x 40 years) you would expect 2.7 with that IQ or higher, so that part seems right.
I think what is going on is that when it says "top 1.33% for the test he took" it is referring to some test given to children who are already known or expected to have high IQs, and on that test people with IQs of 145 are 98.67 percentile.
I agree with the first part but question the second.
It appears that the author was off by precisely one order of magnitude, having mixed up .13% and 1.3%. It could be chance, but I think it’s probably a simple mathematical error. The author's conclusion (2-3 children over a career) is correct, but the intermediate step is mistaken.
I think what is going on is that when it says "top 1.33% for the test he took" it is referring to some test given to children who are already known or expected to have high IQs, and on that test people with IQs of 145 are 98.67 percentile.