Don't see any reason not to consider them "relative" i.e highest peak is 100% rest will follow. Graph needs to be "internally" consistent. No different scale for each nm..
No not "quantifiable" in terms of gross photons.
2 graphs that are the same relatively speaking (0-100 scale) can be quite different "par wise" but its color rendition should be equal.
Problem is "assuming" too much from a relative curve. Say you dim or shut off the predominant channel the new relative spectrum will look like you mega increased the remaining nm's.
RGB..1:1:1
RGB with 50% blue instead of 100% 0.5:1:1
I didn't increase red and cyan..but if I over-layed the 2 and said that the one is def. more reddish green well that would be the likely look. There is not more cyan or red in the second one.
So yes technically not scientific and mainly good for comparing looks but it has other functions.
If the chart shows peak as 120 then one could "shrink" it by diving each nm by 1.20 to keep consistency and overlay. Shrinking in this case not stretching.
To quote AI...
YMMV..... ;)
https://biochambers.com/pdfs/spectral.pdf