RMAN-06023 When Restoring Full Database (new Datafiles)

I'm testing out various RMAN restore/recovery scenarios for 10g R1. Took a full level-0 backup of one of our larger databases (16 tb) and I'm trying to restore the whole thing on another server. While the backup was running (for about a day & a half), a few datafiles were added. The backup completed fine, including "backup current controlfile;". Now my restore is failing with "RMAN-06023: no backup or copy of datafile xxxx found to restore", with reference to the 4 files that were added while the backup was in progress.

I guess if the datafiles were added after the backup had finished, we'd be fine because they'd be created during recovery with our archive logs (we have all of those).

I found some suggestions for first doing a "restore controlfile;" -- did that and it worked fine -- then mounting the database and creating the datafiles manually. I added that to my RMAN restore script and tried it, like so:

run {
allocate channel d4 type disk format '/BK01/dbX/stream1/df_%t_s%s_s%p' ;
sql "alter database create datafile ''/uXX/oradata/dbX/datafile_58.dbf''";
restore database ;
}

The datafiles were created successfully, and the restore began .. but then it failed with this:

creating datafile fno=4216 name=/uXX/oradata/dbX/datafile_58.dbf
...
RMAN-03002: failure of restore command at 06/20/2013 17:27:28
ORA-01119: error in creating database file '/uXX/oradata/dbX/datafile_58.dbf
ORA-27038: created file already exists

So RMAN seems to be telling me that if these 4 datafiles don't exist in the backupset, then RMAN won't automatically create them. But if I manually create them first, then RMAN will try to restore them and complain that they already exist. And for that matter, I would have expected that RMAN would be smart enough to detect that those 4 files were added while the backup was active, and just include them in the backup. Maybe there's something I needed to do to tell RMAN to do that?

This is just a test, so of course I can redo the backup from scratch and try again, but we have some pretty large databases which cannot be shut down for cold backups, so it's entirely possible this scenario could show up in a real disaster recovery. RMAN does such a good job of managing things.