Brian, in addition to Alan's comments about the lack of significance at those very high frequencies, the measurements are also subject to extreme swings with small changes in position. For example, look at the current Axiom graphs for the M3 and M80, which use a "listening window" measurement over some unspecified range of angles. The averaged response at the top doesn't show that drop, and this certainly would also apply to the M22 and other Axiom speakers using that same tweeter if their listening window was shown. Presumably the M22 Axiom spec is based on some averaged number rather than simply the on-axis anechoic measurement.
