I missed the "key" to all of this in the original post. This IS a database that contains historical "Log" tables almost exclusively, right?

If so, I wouldn't do it the way that you propose because making even an empty copy of a problem is still making a copy of the problem. If it were me, I'd invest a little more time and partition the tables (either a partitioned view for Standard Edition or a partitioned table for the Enterprise Edition). That way, you can automate dropping off previous months in milliseconds rather than having to go through the small bit of hell that you're going through now. If you also take the time to make each partition (using either method) on it's own file in it's own filegroup, then you'll also be able to just backup the current month each night instead of backing up the whole shebang every night.