My suggestion would be to endode the PDF in an proc somewhere prior to the xlate and then just map the field through in the xlate. The reason for this is the binary data could cause issues within the xlate.
Let me know if you have more questions.
Steve
