There is certainly a lot of mis information and just plain bad guessing when it comes to the X-Curve for cinema.
Virtually every room used to dub films are very carefully setup to be very close to the true SMPTE X-Curve.
The process used to achieve it does vary, but in most cases, the systems for the screen channels are Bi, Tri, or even Quad amped with sophisticated speaker management systems to correct the individual drivers first, and time align them to get the best possible response with the least amount of Eq to achieve the result. Using continuous pink noise as the test signal is still the most common practice. Many other systems have been tried and there is some merit to transfer function and sweep setups, but the jury is still out on if the results are as consistent. The response measurement for pink noise is done with a spectrum analyzer and multiple microphones. Most systems still time multiplexed the mic's into the analyzer and a long average is used. 4 or more mic's does a very good job of cancelling out peaks and dips caused by room reflections. The system will fall apart if the room reverb is unusual or excessive. There is no easy electronic fix for bad acoustics. Thankfully, most dubbing rooms have very good to excellent acoustics and this measurement system works very well with consistent results. Newer analyzer software and hardware is allowing up to 8 mic's to be analyzed simultaneously and averaged in real time. Not only does this give a much faster true average reading, the results seem to be more accurate because the time of the readings are in sync.
On a dub stage, the tuning setup is a little different than for a large theatre. First off, there is actually a series of X-Curves that vary the high end roll off based on the room volume to reduce the effect on the perceived sound when the mix is later brought into the larger room. Also, in a large theatre, the mic's are spread over a fairly large area to try and average as much of the prime listening area as possible. Before any Eq is applied, all of the mic's are checked to be sure the response is consistent and that no one mic is in an especially poor seat. Checking the mic's in the off axis areas can also show if speakers need to be aimed better. It is very common to use the fall off of the speakers dispersion pattern to help even out the room coverage from the near and far seats. The closer you get to the speaker, the further you go off axis. This can greatly reduce the level difference from the front to back in a large theatre. On a bud stage, the speakers are pretty well aimed directly at the mix console position. The analyzer mic's are usually placed across the console operator positions. The sound out of that area is not as important as what the director and mixer are hearing. The SMPTE spec calls for +/- 3 db from the X-Curve, but we usually achieve much better. In a well equipped and acoustically correct large room, the average is usually even better than +/- 2 db and some can get under +/- 1 db without excessive Eq. On a dub stage, the average of the 4-6 mic's across the console will first be dialed into better than +/- 2 db, usually 1 db, but then we also look at each mic independantly to make sure no one person at the mix console has an error of more than 2 db anywhere. Having the average stray off perfect is better than the director having a 4 db dip in the dialog range.
The high end roll off has been a target of a lot of debate. Many are arguing that it kills the possibility of ever getting true high fidelity as it limits the headroom of the high frequency range with so much roll off. This simply is not the case in the real world. Let's assume a room was tuned to perfectly flat out to 16 Khz and calibrated to 85 db at -20 dbfs. A sine wave at 16 Khz could then hit 105 db at track clipping. This is just utterly not needed. In a smaller room, the response should hold the 3 db per octave roll above 2 Khz. In this case, the level at 16 Khz will hit PCM 0 DBFS clip at 96 db. That is still a sound I never want to hear. Then there is the reality of even being able to punch that much sound into a large room. Very few speakers could handle it for any length of time. There was a track on a feature I will not name, that had a sound effect that did hit close to 0 DBFS at 15 Khz for about 1 second. There were blown HF drivers all over the country. And this was with the roll off of the X-Curve. Without the X-Curve, that signal would have been Eq'd out 10 db louder still. How did that signal make it out?? Probably 2 things happened. Maybe the dub stage was not Eq'd to hold X-Curve all the way to 15 Khz, and maybe the sound engineer's hearing was a little deficient at that frequency. Maybe the director was asking for this short HF burst to have extreme impact. In any case, the system on the dub stage was able to handle it without failing. so it did not send up a red flag of disaster. And there is also a good chance, the system ran into a limiter, probably in the power amp for the tweeter. This could have cause the sound to just seem too low, making the mix engineer turn it up even hotter. The Myth that the X-Curve is killing too much high end is just not based in the real world.
Any good dub stage is calibrated accurately, and the talent creating the soundtrack is laying down the sound while listening to it on this X-Curve, compensated a bit for room size. The tolerance for the X-Curve does allow more roll off at both the high and low frequency limits, but if the speakers are capable, the roll off does not need to be Eq'd to drop off faster. The normal practice is to just stop boosting when the drivers start to roll off. Pushing drivers past their designed frequency limits is always a bad idea. All of the better systems we work with usually have either a few bands of parametric Eq for each driver, driver correction FIR filter, or Dolby LAKE filter bank correction. These can all apply very smooth correction to flatten each driver. The overall graphic Eq that is used before the crossovers on most systems are limited to no more than +/- 6 db of Eq in each 1/3 octave band. If there is a large swing between bands or if more than 4 db is needed, we usually look to see what is wrong. Eq is used to shape the response, not whole sail change it.
I have tuned many small rooms, some quite a bit smaller than any dub stage, and i still use the small room X-Curve. The result is almost always very pleasing on a wide variety of program material. CD's, SACD, DVD and Blu Ray, and even broadcast TV. If a room is especially small and acoustically dead, we will use a little less roll off. In one room I tuned recently I used only a 2 db per octave roll off that started at 3.15K instead of the normal 2K. The room was actually too dead and needed some life dialed in, but it still actually fell within +/- 3-4 db of true X-Curve and sounded great. Pro Logic II music mode added enough natural sounding reverb to make up for the totally lifeless room.
A good sound engineer does need to adjust for the conditions. We have not found an analyzer system yet that can correct for odd rooms. Systems that use ticks or fast sweeps and ignore the room reverb can have excellent results, but the effect of the room will still effect what the listener hears and the results just do not seem as consistent yet.
Any feature film with a reasonable budget is being mixed in a room that is properly tuned to the X-Curve with fairly tight tolerances and is usually brought to a few larger theatres to make sure it translates well. The bigger studios actually have dub stages that are close in size to the average 300 seat theatre to get a better feel for the final result. I know at least one case where the mixer even sat up on a ladder to simulate being in high stadium seating where the back wall surround would be much closer.
When "near field" home video mixes are made, I am sorry, all bets are off. When testing the final setup in a room I tuned recently, we ran several cinema DCP packages and the client was incredubly happy with the results, but when we switched to the consumer source and played a few Blu Ray disks, the sound was all over the map. There was well over a 15 db difference in the dialog levels between the disks. It was clearly obvious that the theatrical mix was not used on at least 2 of them. The frequency response was a little different, but that was not the big problem here. This person wanted to be able to listen at "reference level" which worked perfectly for the cinema DCP's, but there was no constant refernce we could set for the Blu Ray disks. Turning on the dynamic range control in the Denon Pre Pro actually made it a little worse as the Dolby True HD disks and the DTS MA disks behaved very differently to the level normalization. In both cases, they did reduce the difference in the level from the loud and soft passages, but one raised the quiet sections more and the other lowered the loud sections more. Even with DRC on, we had to play 2 of the disks a full 10 db different in level on the Denon pre pro to have the dialog levels match.
In the end, SMPTE has been looking into any better way to tune rooms for more consistency etc. and they did discuss tuning flat vs X-Curve and other curves. In the real world, X-Curve is working very well. Tuning using FIR or filter banks, with sweeps or even music with a transfer function analyzer can all work, and is being used, but they all are still targeting the X-Curve.