I understood what you were saying the first time, but I am also saying that in this case it does not matter because the XLS 2502 has SINAD must lower than that of the 3500's preamp, in fact also lower than that of the power amp. Your logic would apply if the XLS 2502 has higher, or at least comparable SINAD, but it is lower, much lower.
AVR-X3500H's SINAD at 5 W was 82.6/82.4 dB
XLS 2502 at 5 W, SINAD was 69.8/71.9 dB
The XLS2502 only need 1.4 V for it to reach its rated 440 W, that means at 1.2 V, it would still output a respectable a whopping 320 W!
And the 3500 could do SINAD 95.2/95.3 dB at 1.2 V.
So unless you really need 320 W, I would leave the sensitivity at 1.4 V. If you set it to 0.775 V, the preamp output would be worse because it would spend more time at below 0.8 V and the XLS 2502 would be noisier because of the much higher gain.
I certainly don't need anywhere near that power.. picked it on a whim as I was getting a really good deal on it (~$350 equivalent) and am still sure I will find some use for it - if not with the X3500 then elsewhere
The speakers are fairly sensitive (91db) but apparently they drop to <3 ohms for a fairly wide band between 80-600 hz (Focal Aria 926)
On a side note, I am not sure where my calcs are going wrong - but at 1.4 V setting, I have had to set the gain knob at max on the Crown but it's still about 5-6 db less than other channels (directly driven from X3500) while testing multi-channel
Set to 0.775V, the LR matches the other channels with the gain knob set to about 65% although I would have expected this to be more around 30-35%
It's not the amp or a defective AVR either - saw something similar with a Marantz SR6010 and PM17 Integrated too where the PM17 o/p in amp mode ( preamp de-coupled) was 6-6.5db lower than the internal amps on the SR6010.