It does seem bright. That said... while RTings is one of my favorite review sites in the world I do not think their frequency response measurements are useful, or even worse, they can be misleading.
They are only testing in-room measurements with unclear procedures. Without an anechoic response to reference it to nor more details about the procedure/distance/placement, etc, an in-room response is nigh useless. That frequency response could be caused by directivity, room interactions, the speaker's tuning, or any of a myriad factors. Without the anechoic response, you cannot know which deviations from the target are good, and which ones are bad. Sometimes it'll work, other times it won't.
For example, here's their frequency response for the Google Nest Audio:
View attachment 117716
Which makes it look like an awful speaker. But here's my spin:
View attachment 117717
Honestly, no clue how they ended up with that response, but I am quite sure it tells you nothing about how the speaker sounds. It's really quite disappointing they're not providing at least a quasi-anechoic on-axis response, seemingly ignoring all the speaker science out there on the importance of anechoic results and the problems with in-room measurements...
Second, note that they tested the Sonos 5 with TruePlay -- an optional feature -- and only in the horizontal orientation. In the
Sonos One Gen 2 review, we also see an elevated treble which isn't present in the
Sonos One SL review, even though the speakers have identical acoustic components (they mention a firmware update might be the cause).
The most likely explanation could be exactly what you've said: In horizontal mode, one speaker acts as a 'stereo unit, dividing the midwoofers and tweeters into left and right units.
In vertical mode, the entire speaker is mono.
When listening, that basically means one speaker has to do twice the work in horizontal mode.
I'm not completely sure why that would affect distortion measurements if you are playing the same signal for the 'left' and 'right' parts of the speaker, but I very much suspect this very likely has something to do with it.
Two Sonos Fives sounded much better than one from my recollection. Trueplay also worked well -- it's essentially performing an MMM measurement and applying bass calibration. And you could also mess with the tone controls as well.