This is just an alternative (and inferior - poorly/minimally documented and with too few data points) way to do what the 2.5D HueSatMap tables do.
No. Stretching the exposure in oddball ways will completely screw up the resulting profile.
In general as long as the raw image itself is well exposed (white patch is close to clipping but far enough to have decent margin, helps to bracket with 1/3EV steps and choose the brightest non-clipping shot) exact levels should not matter. Adjusting the exposure slider in RT is only useful if you want the image to be easier to view in intermediary processing steps (Hugin for defishing, GIMP for pulling up the pixel coordinates of fiduciaries).
There’s currently a pull request that will improve the dcamprof workflow a bit, but it’s blocked due to rawpedia being in a sorry state for the documentation side of things, migrating rawpedia to something more maintainable has stalled for various reasons - An update on Rawpedia