It doesn't have to, but shifting the y-axis to exaggerate differences is like the most textbook example of deception in statistics. In that sense, if you want your work to be taken seriously, especially when it purports to be more accurate than other statistics, it's certainly a choice.