Apparently, when parsing an (X)HTML which contains an , this non-breakable space will become a regular space in XDOM (and therefore in XWiki syntax if serialized after).
non-breakable spaces from (X)HTML should be kept as non-breakable spaces, since they're probably needed if they were explicitely added as non-breakable spaces.
Note that there are many places where non-breakable spaces are necessary, although nobody actually uses them (e.g. when writing big numbers with separation between the digit blocks, that is non-breakable, e.g. 10 000). The parser should allow correct handling of these cases.