[zanox]

  Grado's are in some ways even more badly behaved than the DT990, but while the DT990 has that big solitary spike at 9k, Grados has a lot of smaller bumps all over the vocal area of the spectrum, which might make it sound perceptually more even.
 
The problem with the DT990 isn't really the spike, all headphones have deviances from ideal, it's more the 10dB difference between 5kHz and 9kHz that's the problem. If you manage to even out that difference using EQ, and add some compensation for your own ears, you should expect a marked change for the better.
 
Burn-in has also been discussed a lot here, and Tyll Hertsens did an interesting test a while back. The consensus from that seems to be that any change that occurs, happens very early on (the driver 'settles' more than it is 'burned in'), and the magnitude of that change, insofar as it occurs, is so little as to be practically insignificant. The simple act of reposition the headphones on your head will cause larger changes in how the headphones sound.

so I should up 5khz by +5 and turn down 9khz by -6 to make it even?
I am a treblehead myself. I love my modified Grados to death. But the brightness on the DT 990 is on a whole new level. They sound really weird to me. There's too much air in the soundstage.  They have this dark airy atmosphere filled with extremely bright sounds. That's how they are sounding to my ears right now. I liked the bass though. It hits me hard. but they are really not so musical like my Grados. I wish I had bought the 325's instead. I skipped on them because people said - they are bass light. What a mistake.

 

[bigshot]

9kHz is a full octave above the vocals. I don't see anything on any of those charts that would account for sibilance in vocals. It's likely just manufacturing variance with those particular headphones. An EQ dip around 3 to 5kHz would take care of it. A little experimentation with Q widths and amounts of correction would locate the sweet spot.

 

[limpidglitch]

Not necessarily that much, I'd do it incrementally.
'Correcting' what you see in the frequency plot us usually a good way to start the process, and being aware of 'pressure points' in your own hearing can be useful as well, but ultimately what sounds good to you is all that matters. Some of the unevenness of the headphone's character might actually be to you liking.
 

  9kHz is a full octave above the vocals. I don't see anything on any of those charts that would account for sibilance in vocals. It's likely just manufacturing variance with those particular headphones. An EQ dip around 3 to 5kHz would take care of it. A little experimentation with Q widths and amounts of correction would locate the sweet spot.

 
I agree, 9kHz is a bit high to be called sibilance, but a sharp peak there can still cause some unpleasantness.

 

[bigshot]

Bringing up 5kHz to match 9kHz is going to make the sibilance worse, not better. You have to EQ with your ears, not charts. The range of deviation just from manufacturing error can be 5dB or more. Only the most expensive headphones have deviation from spec less that 3dB.

 

[limpidglitch]

I'd bring up the 5k band (and down the 9k) so that I can start with something resembling a blank slate. Masking effects has a tendency to confuse me if I don't do it like this.
If I end up making a valley in that area later on, so be it.