In XML, encoding is often provided inside <?xml ..?> tag. In general, you cannot read the tag, if you don't know the encoding, but XML Parsers support the detection of several encodings which allows them to read the tag. With that information they can read the whole file without knowing the "charset" in first place.
xtokenize and xslt use XmlInputFactory#createXmlStreamReader(Reader). But by providing a reader Camel tells, that it knows the encoding, so it won't be detected by the XML parser.
Also Camel sets the charset to UTF-8 if it is not provided inside a header. This makes the underlying reader fail reading UTF-16.
Using XmlInputFactory#createXmlStreamReader(InputStream) inside XMLTokenExpressionIterator works (tried in a patch). But the next xslt steps fails again because it again uses a Reader.
See Stackoverflow Question for reference: