They do not need to "lie," just report the measure of skewness which tells you how far away the mean and the median are apart and in which direction .
Reporting means is advantageous in many circumstances, especially when you are testing for between-group differences in samples (or do regression modeling) which is a means-based test that also takes into account the dispersion (variance). Thus, one can find a difference between groups that have the same median.
Wojtek