Yes the actual text in the document is in document.xml. From what you say, I think you have copied too much data for getting the isolated compressed part. Remember the "Extra field" (if used) are right behind the name (in your case word/document.xml), and is not part of the compressed data. Then remember that a new local file header (for the next file) is located before the next name (in your case word/theme/theme1.xml). That's 26 bytes before the start of word/theme/theme1.xml. Lastly, to decompress that compressed part, you must use a method called inflate, and most importantly you must inflate using the raw method.
If you're interested I can give you a link to the alpha version of my program.
Joakim
My tools are now released here; http//www.forensicfocus.com/index.php?name=Forums&file=viewtopic&t=7814
Any feedback and discussion should continue in that thread.
@Georgefan
I have sent you e-mails. If they did not come through, then send me a message to the adress I gave in the pm.
Joakim