Unfortunately Jeff, I don't have the system I used for this anymore, refurbished it and donated to someone that didn't have a laptop.
The gist of it is that I had some relational data (4NF) exported to XML files, which I then checked for the actual data length. Then, in a vanilla database, I imported the files into clustered index tables as XML data type, using small int as the clustered key (minimizing the impact). Then in a separate vanilla database, I imported the data from the SQL XML storage into the same schema as the original data originated from. Both SQL databases had autogrowth set to a minimal size and the storage quantification was based on the file size deltas, before and after the data import. I know it's not 100%, but I think its close enough.