Assessing bias in total mercury results after removing a subsample from the bottle