The xml field (with or without OE) is not designed for this task. We do the same as you and basically have dropped the idea of preserving: we now ask those MsWord producers not to format too much and our secretary does the (limited) reformatting. Painful sometimes :-(
The xml datatype is even serverly limited in text structure, for example nested lists are not possible.
You may want to revise your approach and let those MsWord things become attachments (binary file datatype) and use the other datatypes to manually copy meta-data.
