First, what tool did you use to measure the frequency response?
But in general, all ML’s have a bit of a high-end roll-off. If they didn’t they’d be shrill. A line source speaker can really over-energize a small room with high-frequencies.
Actually, I believe one of the reasons so many people like ‘stealth tone control’ cables (that roll off highs due to capacitance) or tube pres and amps is that they all contribute to a softening of the highs. But that just my biased view
Also, if you are measuring a post-Audyssey corrected response, with the ‘normal’ Audyssey EQ in place (vs Audyssey ‘flat’), then you are ‘seeing’ the standard Audyssey target curve, which does feature a high-end roll off.
I’m not surprised that you have -14db total drop-off after 13Khz.
My system is also measures >10dB down at 15Khz and over 18dB down at 20Khz. Yet the balance is just fine.
If it weren’t for the room treatments everywhere dampening the resonances, I’d feel they might even be a bit bright (on Audyssey normal). On Audyssey ‘flat’, they’re noticeable brighter.
Also, are you using the ‘current’ source output of the Sunfire? If so, that might lead to that drop off as well. I’d recommend using the Voltage source, re-doing the Audyssey setup (including varying the height of the mic in at least two of the measurement positions to better capture the line-source pressure field.
I could go on but let’s get the answers to the above questions first.