The case for a map-edge pair is fairly clear. Assume there are two American map-edge hex pairs, ( GG2, GG4 ) & ( GG8, GG11 ), the hexes between them are also American and all the rest are German map edges. The two edge pairs are segments separated by three German hexes. This is an arbitrary separation, but I think some of the phrasing in 3.6054 is not as tight as it needs to be for zero hex separation. In 3.6053 the both edge hexes get a pair of perimeter markers pointing at each other.
In 3.6054 you start placing non-map-edge perimeter markers.
The case without attempting two edge segments: you start by placing a perimeter marker that points at GG2. You continue on placing perimeter counters without going through GG8 or GG11, each one pointing at the last, until at last you reach GG4. You have formed a loop. That is a successful completion. Repeat for GG8 & GG11, this time without crossing the previous loop. Successful perimeter.
The case with attempting two edge segments: you start by placing a perimeter marker that points at GG2. You continue on placing perimeter counters, each one pointing at the last, until you reach map-edge hex GG11. This is not a successful completion of a loop, so you have to continue by placing another perimeter marker. If you place the perimeter marker in GG8 then you have violated 3.6054, "the new (Alternate) Hex Grain may not overlap another friendly (Alternate) Hex Grain (even on the map edge)." So you can't continue out of GG11 to GG8 and form the loop containing both map edge segments.
Where this gets a little hinky is either with pairs of map-edge hexes that are separated by zero hexes and with singleton map-edge hexes. The text I quoted above actually continues with an EXC: "the new (Alternate) Hex Grain may not overlap another friendly (Alternate) Hex Grain (even on the map edge) [EXC: in a hex already containing a friendly Perimeter marker]." In my unsuccessful case above I separated the two map-edge hexes with three hexes without perimeter counters. If there were a close pair, say ( GG17, GG18 ) or a singleton ( GG20 ), placement of a second perimeter marker from GG17 to GG18 would result in two hex grains that overlap but that seem to meet the EXC. In the cases of adjacent pairs or a singleton, it might be said the placement meet the EXC. Based on the fact that segments separated by one or more hexes are not allowed, I am guessing that close pairs and/or singletons are not allowed either, and that the rule is twisted and malformed
As for forming a question for a q&a, I would suggest using examples as above rather than trying to convey what you mean in the abstract.
JR