Here are a couple of things that I routinely deal with:

1) If you routinely refresh data in a test server, it's much easier to restore a database than to more over just some of the tables. The former requires scripting out security, the latter requires scripting out foreign keys, and dropping them first before the refresh.

2) If different backup cycles are needed for different portions of the data. And if some of the data is static, then read-only filegroups would save on backup space but make for a more complicated restore.