The Circularity is Near?

The ILR/ACTFL scales have been described as criterion-referenced by their proponents. However, this claim has been disputed on the grounds that they are not truly criterion-referenced because of the built-in circularity of their use. This circularity is usually described as such:

How do we know a speaker is Advanced? Because s/he can do Advanced level tasks. How do we know the tasks are Advanced? Because they can be done by Advanced-level speakers.

Criterion-referenced test development is premised on the mastery / non-mastery distinction. To a certain degree, don't all descriptions of mastery start with an identification of people who would be considered masters? And wouldn't one logical starting place to identify those "master level" tasks be precisely those tasks which "masters" can do?