Measurement mics are all calibrated for 0 degrees incidence, and none have equally flat grazing response (90 degrees), but the differences are confined to the extreme high end, the point at which the zero an 90 degree responses diverge is diameter dependent. For a 1/4" diameter mic, expect -4 to -6dB at 20KHz at 90 degrees.
For the "good stuff", like Earthworks or the best, Bruel & Kjaer, the grazing response characteristics are known, and can be integrated into the measurement system for random field applications.
90 degrees "grazing" orientation is used for measuring moving sound sources like traffic noise, and in that application there's no need for perfectly flat response at the extreme high end, though obviously correction is possible. But that orientation is also useful for random field measurements, like those we'd want in a room.
For precision high frequency measurements, the mic must be pointed at the source. However, that's not really what we do for home theater measurements, and for some measurement types grazing orientation would be more useful, such as looking at the spectrum of a reflection vs the direct spectrum. We'd technically get more meaningful waterfall plots using grazing, though we'd have to compensate for the slight high end loss of the direct signal for the plot to be completely accurate.
You'll notice that measurement mics included with AVRS (like the Audyssey mic) are meant to be used in the 90 degree grazing orientation, because they are working with the entire sound field, and are calibrated for use that way.
In practice, grazing is probably more useful when looking at speakers in a room, but it's a fine point because most of our serious EQ work happens well below the zero vs 90 degree response divergence point anyway.