Adding features that are not part of the cell could "break"
feature-based recognition in the extract deck. Shorting
substrate and an inductor terminal might cause pruning,
depending on rules setup. Might begin with removing the
P+GR feature and see what occurs. Might look at extract
deck text and see what the logic for recognizing this
particular inductor is, particularly what adjunct features
(say, "inductor/drawing" polygon in an AND logic) are
expected / required. Maybe drawing the recognition
layers, is on you (would be a sloppy library development
choice, but can't be ruled out).
Not clear to me whether the complaint is from the layout,
or that the inductors -are- found but are missing (or some
other, not-same primtive is in the schematic instead). You
have to look at both schematic and layout perhaps; this
report format is unfamiliar to me.
Go to the extracted view and turn off selectability except
for instance; area-select an inductor and see if you can
find the itty bitty extracted instance that gets placed on
the extracted layout, and property-edit. That will tell you
what (if anything) the inductor as-laid-out is being "seen"
as.