Wow, indeed. That shows a nearly 12dB spike at 9.5 KHz. Yeh, if that's accurate then I'd definitely say it qualifies as "hyped".
The differences between the two measurements can be chalked up to variances between tested headphones and smoothing. Smoothing refers to the number of FFT bins, or width of frequency bands reported. A graph with no smoothing is almost useless because it shows too much detail, much of it irrelevant. At the other extreme, too much smoothing hides imperfections, a common practice among marketers to make their products look better than they really are.
Just one of the reasons frequency-response charts are to be taken with a sodium chloride crystal, and best used to reveal general observations rather than as the sole determiner of quality.