Not sure if you have seen this, but it seems like a reasonable review: http://www.inearmatters.net/2010/12/review-hifiman-hm602-and-hm801.html.
In particular this excerpt from the article seems like a reasonable explanation to the roll off, and the reasons behind it:
"After HifiMan released, it has been met with a lot of doubts and criticisms. Some of the more noticeable one is the treble roll off (attenuation) between 10 kHz to 20 kHz. It is about -2.5dB on 15 kHz and roll down to about -5dB at 20 kHz. While some consider it as a defect in design, it is actually being done intentionally. There is always a low pass filter after the DAC stage (or integrated internally) to remove the high frequency component (> 20 kHz) in order to remove noise as well as the inaudible frequency. Human hearing is generally recognized as 20 Hz to 20 kHz, but in real life the full range is between 16 kHz for most adult and down to 80 Hz in bass (*anything below increasingly tends to be felt more by skin rather than heard by ears, which we often refer as ‘chest pounding sound’). What HifiMan employed is called a Butterwoth filter (a.k.a. Maximally Flat filter). AS I have read, the characteristic of this filter is that the passband (the desired range) has a maximally flat response (no ripple in frequency response) as well as a more linear phase response (no uneven phase shift in individual frequency). The downside is that it doesn't roll-off quite as fast as other types of filters (which create the slope or ‘roll off’ in HifiMan’s upper treble). I think the consensus is that there is no such thing as one perfect filter for everything. Each type of filters has its own pros and cons and it is up to the implementation as well as what goal the designer wants to achieve to determine what is best for a certain design. In fact, quite a few well regarded high-end DAC and CD player also employs Butterwoth filter because it is considered more analog, musical sounding than other filters. I think the problem of Butterwoth filter on HifiMan has less to do with actual performance, but whether some can accept that you don’t need absolute perfect flat line to create a player that sounds good. If you can’t, then you know what you won’t be buying. I am not an electronic or audio engineer that can tell you what should have been done that could yield a better result or whether a Butterwoth filter is indeed the right choice – but rather trying to interpret whether the end result sounds good to me or not"
The compromise here is an early roll off in order to get flat frequency response and linear phase bellow 10kHz. The roll off is close to -2.5dB at 15kHz which is just beyond my current hearing range so I don't think I will miss it.
Pay particular attention to the "Navigation and Format Support" section. Don't expect the navigation to be on par or as advanced as the one that comes with your iPod.
Another important point is that the HM 801 seems to be 18 ohms while the P5 is about 27 ohms (http://www.innerfidelity.com/images/BowersandWilkinsP5.pdf) It may not be a problem as the P5 seem to have a fairly flat impedance chart, but the two will likely interact and roll off the high frequencies of the P5 a little bit more. It has been shown to do so on certain headphones (http://rmaa.elektrokrishna.com/Comparisons/Hifiman%20HM-801%20-%20Impedances.htm)
More impressions here: http://www.head-fi.org/t/563735/hifiman-hm-801-sounds-no-better-than-my-ipod-touch-3g
(don't let some of the comments discourage you from it though. Measurements are good indicators of performance, but not the last word IMHO)
Edited by ultrabike - 11/26/12 at 12:21pm