That's a great reference, and a sensible discussion of the essential arbitrariness of the definition.
In experimental modal analysis we often take the modal mass at the point of excitation - so we know that if we add 5% of the modal mass to that point we'd expect a -2.5% shift in frequency (or at least, on a good day, that it will drop a bit!). What seems absurd is that by this definition the modal mass varies if you change the excitation point or direction, but that is correct - adding mass at a nodal point will have little effect, whereas adding it at an antinode will have a great effect. Cheers
Greg Locock