I'm using both the format query parameter (SPARQL standard) and the HTTP Accept header (HTTP content negotiation) to specify my desired results format (application/sparql-results+xml) and it only returns the full web page with an embedded html table of results. Is there any documentation on this behavior?

Here is the request header included in the response from the endpoint for one of my queries which seems to indicate that the Accept-portion of the request has been properly understood (and ignored?):

   GET /sparql?timeout=0&format=application%2Fsparql-results%2Bxml&query=PREFIX%20obo- term%3A%20%3Chttp%3A%2F%2Fpurl.obolibrary.org%2Fobo%2F%3E%0ASELECT%20%3Flabel%0Afrom%20%3Chttp%3A%2F%2Fpurl.obolibrary.org%2Fobo%2Fmerged%2FGO%3E%0AWHERE%0A%7B%0A%20%20%20%3Fannotation%20rdfs%3Alabel%20%3Flabel%20.%0A%20%20%20VALUES%20%3Fannotation%20%20%7B%3Chttp%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FGO_0005829%3E%20%3Chttp%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FGO_0005886%3E%20%3Chttp%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FGO_0005654%3E%7D%20.%0A%7D%0A&debug=on HTTP/1.1

   Accept: application/sparql-results+xml
   Host: www.ontobee.org
   User-Agent: HTTP.jl/1.4.2
   Content-Length: 0

The response (which I am omitting for brevity) includes the desired results for this query but they are embedded in the source for the full human-readable page, how can I just get the xml?