There are often layers with no physical meaning, that the
CAD guys use for recognition, tagging, keepouts and so on.
And a real mask layer may be built from some superset of
layers / blocking layers / purpose-layers by Booleans and
operators (over/undersize/avoid). So it's not surprising if
you run across things that are not intuitive or have no
obvious 1:1 meaning.
What I've found a useful learning exercise, is to place and
flatten a "normal" device of each type. and then "roll it"
from X-Y to a X-Z orientation by stretching the polygons
to a roughly-right Z height from their original Y extents,
and build myself a cutaway view. You start with the more
obvious ones (well, active, S/D, poly, contact, metal) and
apply some mental Booleans of your own (like, (poly+!active)
hard-masks S/D to less than implant PR) and make your
device look right, classically. Then you will have these
"leftovers" to figure out, but now you have the context of
the basic device that can show you, perhaps, that C1 must
be contact because there is no other contact where you
need one, etc. (just an example).
Your foundry ought to publish, somewhere, what the mask
Booleans are in terms of PDK named layers. You can look
through those and find where "whatever" is being used,
and by its use deduce what it's meant for in layout. But
for some people, a visual view is more helpful, which is the
point of what I suggest above.