I believe the paper (or rather, perhaps, the article) makes a wrong conclusion. It doesn't control the fact that US hosts a relatively much larger number of scientific research compared to other individual country, due to its ability to raise funding. This inflates the proportion of US scientists.
Yes, if more than 1/3rd of the scientists submitting papers are Americans, than American scientists are guilty of fraud LESS OFTEN (per person) than scientists who are not American.
Having been involved in science for awhile now, it is perfectly reasonable to assume that over 1/3rd of all work being done in science is being done by an American scientist.
That’s a good point. However, the article does address that US scientists are the most likely to be repeat offenders.