Avamar slow for file level restore? (and missing attributes, on linux)

Hi all, testing out Avamar in the EMC lab.  Having read a laundry list of limitations in vmware tools API-based restores on vmware's site, I wanted to do some additional testing.  The EMC lab is set up with a virtual data domain as the target, avamar doing VM-based backups.  One of the vmware limitations listed was inability to restore more than 5000 files at a time, or possibly in a directory.  I created a directory with 10,000 small text files, each with unique content, and varied the ownership and permissions of some of them.

I ran a backup and began a file level restore of the test directory.  The rate that the files were coming back was pretty poor.  It would have taken about 12 hours to do the whole restore; I broke it off after three, so I didn't get to the point of hitting the 5000 file limit.  Additionally, on the files that had been restored up to that point, all of their meta data was missing, including last modification time, ownership and permissions.

Is this a limitation of file level restore through vmware tools?  Any workaround for these issues; either the speed problem or the loss of metadata?  Our company does a LOT of single file or single directory restores each day based on customer requests.  We currently use netbackup for this, but doing the file level restores there require a netbackup agent (just for the restore, still doing vm backup), so they're fast and retain the metadata.

Thanks!