Yep, it's clearly exposing for the sky rather than forground. From what i've read, there isn't much headroom for highlights before they get blown on these x-trans sensors so it might be that fuji have configured their software to do this.
The fact is though, that there is a live view and histogram to show you, and a simple ae lock button, and a big twiddly exposure comp dial.
I imagine spot metering will fix it for you though. I shoot much the same, I meter and focus with a small central focus point, and then compose.
It is a misonception to think that xtrans sensors are short on headroom. There is loads to recover from the raw files , both for shadows and highlights. However all jpegs by their nature and design have a more restricted brightness range, and the top and bottom of the range is discarded with all the "surplus" data. It is common knowledge that the jpegs fom the X#0 series of cameras respond well in pp to a quite substantial adjustment.
on the matter of exposures, all digital exposure estimation, just like indipendant exposure meters. Work on the basis that the average luminance of the whole of the scene adds up to the equivalent of an 18% grey (actually nearer 12%)
This is also the way spot metering functions, as it sets what it "sees" as that 18% grey, what ever its actual tone might be.
a majority of settable cameras offer at least three ways of measuring exposure. Spot...average.. and multi
Average is just that, it measures the whole of a scene and gives a slight preference to the centre area. Creates an exposre equating this, as if it added up to 18% grey. Multi, is rather different as it uses algorithms that compare the exposure pattern that it sees, with a database of scenes,
it uses these to help compute an exposure.
None of these methods are as accurate as using a hand held incident light meter.
Cameras like the X#0 series are not point and shoots, when used in the PASM modes they are operationally almost identical to a dslr.
However just like lower end dslr's, they can be set up using the various auto and scene modes to be virtually fully automatic.
I have been using adjustable cameras for some 70 years, both as an amateur and through a working life as a professional.
Even now I would not expext my shots to fall outside about 1 1/2 stops if the camera was left to itself. If I take control myself, I would expect that to come down to 1/2 a stop from what I expected. If I were to use my incident light meter I would expect the exposure to result exactly as expected.