Yes, thermal expansion of the bed at different temperatures is enough to throw out the first layer height.
The built in calibration routine heats the bed to 60C, so if you print at 90C bed the gap would be slightly different, however the Z-axis homing that is performed at the start of the print is done after the bed is up to temperature so this should be accounted for - at least in the centre of the bed.
If you did most of your printing at 90C you would probably want to do the mesh bed levelling at 90C as well, as the expansion between 60 and 90C may not be even across the bed - there is no way to do that in the touch screen interface but it can be done via issuing G codes over the USB port. (For example using Octoprint)
In my experience I’ve found the Z-axis offset adjustment (and by extension the mesh bed levelling) THE most critical thing to get right to prevent prints failing.
It needs to be correct within +/- 0.05mm. Even 0.1mm out in either direction will cause obvious problems with the way the first layer is applied.
Too high and you won’t get good layer adhesion causing the print to lift as it prints, (sometimes even parts of the first layer won’t stick) and too low will cause a lumpy/bubbling first layer with visible signs of the nozzle tip scraping the surface.
I usually start with the paper test (standard A4 paper is about 0.1mm not 0.2mm as the calibration sheet suggests!) to get it close and then do a number of small prints that are flat and just a few layers high and watch as the first layer goes down to see if it lays smooth and flat or starts bubbling or scraping, then adjust in 0.05mm steps until it’s printing right.
I print this little thing that I designed for doing this first layer calibration:
It also has a couple of small towers and bridges to help check stringing, part cooling and bridging performance, and only takes about 4g of material and 15 minutes to print.
Measuring the gap with paper or feelers is difficult because there is usually a tiny blob of filament stuck to the tip of the nozzle if you have printed recently, this is enough to make the gap seem smaller than it really is unless you make sure you clean it away, and also remove the filament so more doesn’t ooze out during the calibration process.
So I do the fine adjustment based on first layer print quality rather than trying to do a super accurate measurement of the gap.