When is consistency is more important than the truth?
When making decisions using data trended over time, consistency in how the metrics are defined is substantially more important than whether the numbers are accurate to the third decimal point. People, and managers in particular, like to know if things are better, the same or worse than expected. Market share of 13.244% is much less valuable than knowing that it is up over last year. To make sure that it really is up consistent definitions and calculations are required in both numbers.
This lesson was learned years ago in the syndicated data world when errors in projection schemes were corrected only when we were sure the relative change was still consistent. Simply having a more accurate number is not always the right answer.