Here and everywhere else in this document, where UTF16 is mentioned, UTF16 big endian is implied. Little endian variant of UTF16 is not supported as it is not supported by XHCS. When files in UTF16 are added/extracted to/from archives, BS2ZIP does not add BOM if it is absent in the file and it does not remove it if it is present.
Both Zip and Gzip archives created on open platforms do not have information about the encoding of files inside of them. This is why by default before extracting a file, added to archive by different archive manager, BS2ZIP scans the first 32 kiB of it to determine the most likely encoding between ISO8859F, WCP1252, UTF8 and UTF16. To guaranty a correct encoding and skip the scan use EXTRACT-FILE with CHAR-ENCODING = *BY-PARAMETERS( ... ).
Files added by BS2ZIP to winzip compatible archives with DATA-TYPE = *NOT-SPECIFIED or *CHARACTER have CRLF in corresponding encoding as line delimiter.