Well, having recently done my apprentice judging at the IPMS/USA Nationals
this past summer, I can tell you that they specifically addressed the
accuracy question with examples similar to what is cited here. The
philosophy is that it is impossible to be fair unless you are an expert on
every subject which, despite some 'expert' opinions is impossible in
itself, so the modeler must be given the benefit of the doubt. That even
includes the offset compass judged by the 190 expert because 1) as you
point out the 109 might have an equivalent or worse accuracy error and 2)
who is to say that there wasn't some 190 that had that deviation.
Certain things which are standard practice such as position and alignment
of national markings do get considered if they are 'wrong' unless the
modeler can provide documentation of the discrepancy.
This is the IPMS philosopy as explained to me. There are other schools of
thought.