Saxon

The XQuery code declares a default collation, which the XSLT code doesn't. It's almost certainly the collation that accounts for the EE/HE difference, because Saxon-EE uses ICU for collation support whereas Saxon-HE uses the native JDK libraries.

The semantics of substring functions in the presence of a default collation are fairly peculiar because they cause certain characters to be treated as ignorable. Unless you really want a collation-sensitive substring, I would avoid this area.

It's worth reading

https://www.w3.org/TR/xpath-functions-31/#substring.functions

for an explanation of what is happening here.

The substring-after function says:

The function returns the substring of the value of $arg1 that follows in the value of $arg1 the first occurrence of a sequence of collation units that provides a minimal match to the collation units of $arg2 according to the collation that is used.

If the second argument of substring-after() is a character (such as '#') that is ignored for collation purposes, then the sequence of collation units for the second argument is empty, which means that it matches at the beginning of the string, which means that the entire string is returned.

Thanks for explanation !