Let's do some weird math with that pic
1080p native = 1920x1080 = 2073600 px
1440p native = 2560x1440 = 3686400 px
1800p native = 3200x1800 = 5760000 px
1440p looks to have 64% of the 1800p px... if we take that with the 12.2ms from 1800p then you will have 7.8ms for 1440p.
1440p looks to have 178% of the 1080p px... if we take that with the 5.2ms from 1080p then you will have 9.2ms for 1440p.
I can't say for sure where 1440p native will land but it is most probably between 7.8ms and 9.2ms.
So for this game even in the worst scenario the 1440p native will be faster than 1800p CBR (9.2ms vs 9.6ms).
But that only accounting for this test render time (I don't know what is he rendering in the test but it probably not a full frame maybe an effect because in 30fps the each frame can be up to 33.3ms... so 9.6ms seems to be some part of the render only)... anyway the point is that there are others things not showed in this test like bandwidth use, CPU user, etc... he choose a part of the render to show his gains in that part... maybe a full frame render should be a better comparison because it englobes everything.
Maybe 1440p native can indeed works.
It just that I don't believe they didn't test 1440p native at all... resolution is not chose like that... they tests several scenarios to decide which one to optmize and use.