It looks like you're using an unsupported feature. Have you read the documentation?
We could use some more "cooks" to help complete the implementation of the memo document replacement feature. The way it is supposed to work is that the original MIME contents of the incoming memo should be stripped out. And then the MIME output from SpamAssassin (including the report) should be put back in its place. That way the user will see the safe version of any spam, along with a report on why it is spam. I got this working for some cases, but it seems to screw up the MIME structure in other cases where there are multiple parts. So if anyone has time to work on this it would really help the project.