The levels seem to have changed between the initial "baseline" tests that you did, and this new set of tests. The new one are several dB quieter. It's hard to say exactly how much, since they aren't the same, due to all the changes you did, but it looks like about 3 or 4 dB difference. Its important that you keep the levels the same, always: if not, then it's hard to compare the graphs.
I tried the sub at the 25% width facing the opposite wall
I didn't see that measurement in the new REW file: Please post that one too. You don't need to do a full set of tests each time, for setting up the sub: just one single LRS test (Left speaker, Right speaker, and Sub all going at once).
but I think the desk might be messing with the audio frequencies,
It,s possible, yes, but probably only with mid and high frequencies, not much with lows. There's a basic principle in acoustics that objects can only have a significant effect on sound waves where the wavelength is smaller than the object. At 100 Hz, for example, the wavelength is 11 feet, so any object smaller than about 11 feet wide won't have much effect. At 225 Hz the wavelength is 5 feet, so assuming your desk is 5 feet wide, it could start having an significant effect on frequencies from 225 Hz up.
so I moved it to the middle of the room against the wall facing to the back wall.
A single sub should almost never go in the middle of a wall, as that means that it will trigger all second-order length-wise axial modes very strongly, since it is at the pressure peak location for all of them on the front wall. And since it is also at the center of the FLOOR, it will also trigger all second-order height-wise axial modes. You'll get a very strong modal response from having the speaker in the middle of the front wall. That's not a good location.
Please try the 25% locations again (both 25% left and also 25% right), and do REW tests like that. Also, at both of those points do a second test with the phase of the sub flipped to the opposite setting.
Locating the sub in the room is a slow process: it's not as simple as just trying a couple of places, then leaving it in one. The process STARTS with trying the 25% locations, with both normal phase (0°) and flipped phase (180°), then moving it in very small steps, just a couple of inches at a time, and testing in each location with REW, and keeping careful notes of where each test was done. Then you can compare all of those, to see which one is giving you the smoothest response in the TIME domain, as well as the frequency domain. It is similar to doing the "walking mic" test (see here:
viewtopic.php?f=9&t=24 ), except that you are moving the sub, not the mic.
I also lowered the speakers a bit so the tweeters are about the height of my ears when listening.
That's probably a bit too low. It's not the tweeter that should be at ear-height (that's a common myth, unfortunately), but rather the acoustic axis of the speaker. Some manufacturers state the location of the acoustic axis in their manuals, others just publish it on their website, and others don't bother. If you don't know where it is for your speaker, then image a line that joins the middle of the woofer to the middle of the tweeter: the acoustic axis will be half way along that line. That's the point that should be at ear height, or slightly above. The "standard" height for speakers in studios is to have the acoustic axis 120cm above the floor, which is 47 1/4". However, it is also common to raise the speaker a little higher than that, maybe an inch or two, maximum (3-5cm), for more complex acoustic reasons.
I also moved the speakers closer together and pointed them 16 inches behind the listening position.
You can do a similar procedure as above for finding the best location for the speakers, in terms of separation: start with them very close together, then do a REW test like that, then move each speaker 1" out (in other words, increase the distance between them by 2", since each speaker moves 1") and to another REW test like that. Then keep on moving them in 2" increments, until they are nearly against the side walls. Then look at all the tests, and find the location that has the smoothest time-domain response while also having good frequency response. in this case, you need to do the L-S and the R-S and the LRS test at each location, assuming that you already have the sub positioned in the best spot, or if not, then do L, R, and LR tests.
Getting the speaker and mix position layout optimized is a long, slow, tedious process. It takes time: not something that you can do in a few minutes with just a couple of moves. When you look at the animated graphs in the "Walking Mic" thread (see link above), you can really see that even small changes in position of the sub, main speakers, or listening position can show large differences.
- Stuart -