How much data did they ignore to achieve this desired result?
I think they use the same accountants and adding machines that the BLS uses for unemployment statistics.