I figured that I'm probably not the first one to think about this problem, so I went trolling the intertubes for ready-made solutions. Since perl is my glueware language of choice, I searched until I found the following handy snippet from prlmnks.org:
use XML::LibXML;
my $parser = XML::LibXML->new();
$parser->recover(1);
my $doc = $parser->parse_file($ARGV[0]);
print $doc->toString(1);
Very, very nice! Now I am part of the way there. Next, I took a pre-existing MS Word document of similar make and model, and prepended it. With a little manual massaging, I got the script above to parse it, and even pretty-print it (a nice bonus). Unfortunately, Microsoft Word still doesn't like the resultant "document."
I'm still working on this problem, but that's decent progress for an hour of work.