100GB isn't really that large...but that being said, can you let us know what that space comprises of? Are there many non-clustered indexes? If there are, I'd recommend you creating a new file group, add a new file, drop and re-create all the NCI's on the new file - this would at least place the space used by the NCIs on a different data file.

If you take further steps (as you already attempted) to split the actual data across multiple files, you will get the fragmentation, but some simple index rebuilds should clear that right up for you