A driver that handles 4.5V to 18V has to use thicker gate oxide in its output devices than one rated for 3V upto 8V, for instance.
Thicker gate oxide means higher thresholds and plateaus, and more variation in both, but also will survive higher Vgs transients.
Datasheets are the basis of any legal dispute about an electronic component, so there are de-facto legal documents (I'm sure most/all
companies have disclaimers about accuracy of datasheets, but they can't just put provably incorrect claims in them and expect
no recourse if they don't fix errors / inaccuracies - that would be fraudulent as I understand it. But only guaranteed values / worst-case
parameters are actually a claim about the device, everything else could be viewed as marketing I suppose, but certainly most graphs
are "typical", which has no actual utility other than interest value - you certainly can't infer two products are the same if the graphs are the
same, they might just have been copy/pasted. You do sometimes see ensemble graphs from a sampling of product to give an
idea of spread, which is definitely much more useful, and sometimes you get min/max bounds graphs, again useful.
And the corollory holds - if you have components fail, but the manufacturer can establish you exceeded the absolute maximum
ratings, you are unlikely to have much luck getting money back.
So a lot of a datasheet has legal ramifications, which is why they usually are pretty reliable (so long as you read all the notes and
caveats, and purchase the actual product, not a counterfeit...)
These days of course you have to use what's actually in stock, I suspect this is going to have knock-on effects in future years when
we discover many products compromised on parts during the chip shortage...