There is a bug/feature in the Bigdata UpdateServlet line 520:
If you use a separate Content-Type header, then the contentType variable in that place will contain exactly the string you inserted in the header. That string isn't parsed. So if you happened to use something like this:
Content-Type: text/turtle; charset=UTF-8
Then the code will call:
... which in turn won't work, because RDFFormat doesn't do content type parsing either. Only Content-Type: text/turtle will work. AFAIU, for proper processing of non-ASCII data, the charset MUST be specified in the overall headers of the HTTP request, as in the example in point 1.