Thanks, Sam. I figured it would be something simple like that.
I ended up skirting the issue; I do start with parsing the token stream
and then using the Decoder.DecodeElement on specific start elements
(for reasons entirely unrelated to the original questions).
Dan