The point is that the dates aren't a 'static' cutoff, but more a general concept of where the cutoff is located - obviously the extreme ends of each range will have some issues.

Also, there has been fascinating research that goes WAY beyond 'marketing' in terms of generational similarities - it turns out there are experiential 'constants' (obviously they're not really constant, but overwhelming throughout the population) that make a difference in areas such as learning style and etc.