I don't really know for sure, but this sounds to me like it might be
related to NFC (precomposed where possible) vs. NFD (decomposed:
diacritics as separate characters), two distinct ways of storing the
"same" unicode data. NFD is the most flexible/consistent, but NFC is
more compatible with the past. In this case (as with the ñ), since a
precomposed character does exist for your combination, NFD and NFC differ:
NFC: LATIN SMALL LETTER G WITH BREVE
NFD: LATIN SMALL LETTER G , COMBINING BREVE
For a quick and dirty check, you could try opening your exported file in
BabelPad and running Convert / Normalization Form / NFC, to see if
Lexique then automatically detects those as single characters. But
ultimately, the correct thing to do would be to configure Lexique Pro to
recognize certain sequences as single characters, just like you would
with a multigraph such as the English "th". You might try seeing if
there's any help in Lexique about digraphs.
Jon