Python Bite Is Without Venom | Dr Dobb's

I don't understand that. 400,000 lines times .005/1000, in my book, equals around 2 defects. Interpreting the 0.005 as the more reasonable 'per line', it would expect 2000 defects, not 1000. Does that mean there are 1000 other known defects in Python that Coverity did not find, or that they found, but ignored as not being new? Or did they check what fraction of known defects they found, and extrapolated from there to get an estimate for the number of not found defects?

Dr Dobbs copy-pasted their press release at http://www.coverity.com/company/press-releases/read/coverity... and the report requires registration, so can anybody explain?