Given simple error factors such as measuring device accuracy rounding errors in calulations, etc, etc, etc, you cannot ever conclusively achieve "exactly scale", that's why I voted for 1% off, although I believe that 2% is probably more realistic for the same reason as stated.