There's a bit of misunderstanding of what the X-curve actually is. For example, you see a FR graph with a 3dB/oct. drop above 2KHz, it looks like the X-curve has been "baked in". That's actually not what's going on.
The X-curve is a measurement technique, not an EQ curve. It was developed so that large reverberant cinemas would reproduce the same sound quality heard in small, dead mixing rooms. The curve arose from an anomaly encountered when measuring steady-state pink noise in a reverberant space, where the reverberant spectrum is integrated with the direct spectrum, actually causing the measurement to be inaccurate. Adjusting a 1/3 octave EQ in a large reverberant cinema so that a 1/3 octave RTA displays the X-curve results in a better subjective match to what was heard in the mixing room.
Simply, it's a way to measure, then EQ a large theater to get it to match a small room. The actual correction applied will be different for different rooms, and of course, different speakers.
To quote Ioan Allen of Dolby, "The target of standardization of monitor characteristic, whether cinema, television, music recording, indeed any environment, has to be that material can pass from one to another without requiring any by rote equalization." (“The X-Curve” by Ioan Allen, SMPTE Motion Imaging Journal, July/August 2006) That means, if all is working well, the same track will play on any size system in any size room with any reverb time and sound basically the same. Unfortunately, that's not quite what happened. The X-curve became a rigid standard everywhere, and thus got misapplied. We ended up with tracks being mixed in a small dead room that had been EQd based on X-curve measurements, which of course, was wrong. Today, most of that has been straightened out.
The X-curve has no place in the home theater at all, because an HT is a small, non-reverberant (by comparison) room already, so match to the mixing room has basically already been done. Further, because the problem X-curve tries to address involves measuring continuous pink noise with a real-time analyzer, you can't replicate the same result using a time-domain analysis system like REW, which by its very nature tends to window-out reverberant spectral build-up from the measurement. The correction of X-curve would still be necessary in a large theater, but a time-domain measurement technique would achieve a different result.
So how could a speaker manufacturer bake in an X-curve? He can't, because he doesn't know how reverberant, or how large the space is. It's not a fixed EQ curve at all, is a measurement that includes the speaker, room, and continuous pink noise displayed on an RTA. You can't predict the X-curve and bake it in because you don't know what room you're correcting for.
Any speaker manufacturer making HT speakers for small home theaters claims to have built-in the X-curve has basically not done his homework, and is doing his customers a disservice.
There are also some issues with the X-curve's accuracy. It has an error built into it caused by the measurement equipment of it's time of origination (1972). That error is the basis for THX re-equalization in home systems, and is also the basis for some of the "Cinema EQ" settings found in AVRs. However, the correction is no longer uniformly required, and as a result, Cinema EQ and THX Re-EQ cannot simply be left on. Unfortunately, there's no tagging on media to indicate if Re-EQ is necessary for the home, so you just have to listen. If the track sounds consistently bright, it's time for Re-EQ.
Audyssey would partially compensate for speakers with the X-curve built in, but probably not well because Audyssey has a maximum gain of 9dB, which corresponds exactly to where 16KHz response would be on the X-curve. If the curve were extended to Audyssey's full 24KHz correction bandwidth, it clearly falls outside of the maximum gain. Even if the curve is extended out to 20KHz, flattening that out would be beyond Audyssey's gain range. There's some doubt that it would fully un-do the X-curve though, because nothing really measures smoothly, so I'd expect Audyssey to "give up" at some point. Audyssey does include two target curves, Reference (basically, a curve to correct for the mistranslation built into the whole X-curve mess), and Flat (music). I'd be willing to bet hardly anyone ever switches between them...
Just to be clear, no home theater should be adjusted to match the X-curve, no speaker anywhere should have an X-curve baked in. And the X-curve is not a target EQ curve, but a measurement technique involving continuous pink noise and a real-time analyzers.