Why is it advisable to run a mixed melting-point determination on samples of two different compositions before concluding that the two compounds are identical?