Maybe their tool got confused with the forks made around that time to fix the copyright issues? Basically the tool counted the codebase 3 times due to forks? It seems weird that an error like that would make it to publication though.