Well I certainly hope that my comment will not be understood as a dismissal of graphs or measurements, far from it.
In fact had I performed in situ, in ear measurements with a method better suited IMO to these higher frequencies it's quite likely that I'd have measured a pretty significant peak exactly in the right spot.
It's just that in some cases (emphasis on "some"), a specific individual's experience may deviate from what is measured on an ear simulator, but that doesn't mean that anything goes either and that this divergence can't be explained... by more measurements (preferably in situ then).
In regards to the Mark Levinson 5909, for example, I don't think that a lot of people are likely to experience the 1.5-4.5kHz range in a way that deviates significantly from ear simulators for reasons a little long to explain here.