I have two problems using include_skipped in Catalan grammar.xml for generating suggestions.
1) When using include_skipped="all", all the skipped tokens come together without spaces in between. See rule COM_SI_ES_TRACTES. Example: "com si d'un heroi grec es tractés". It becomes "com si es tractés d'unheroigrec", and should be "com si es tractés d'un heroi grec".
2) When using include_skipped="following", if there are no skipped tokens in the phrase, then it appears the first (the main) token, but I expect no token to appear at all in the match. See rule COM_CONDICIONAL. The phrase "Com no vinguis, m'enfadaré" is corrected as "Si no véns, m'enfadaré", which is OK. But the phrase "Com vinguis, m'enfadaré" becomes "Si Com véns, m'enfadaré", and should be "Si véns, m'enfadaré".
This problem has been solved by preserving the isWhiteSpaceBefore data in Match.java, method filterReadings.
Seems to be fixed according to Jaume's comment, closing the issue.