Mean grain diameters from thin sections: matching the average to the problem