If you set up a witness, you should be able to get the secondary running when the primary goes down for a long enough time (the witness has a timeout value). Using FORCE_SERVICE_ALLOW_DATA_LOSS will allow you to get the mirror back up, but if the primary is only isolated, and not actually down, it is basically breaking the mirror, as both replicas are now on their own individual LSNs. have a look at the cautions and caveats on the Technet article:

http://technet.microsoft.com/en-us/library/ms189977%28v=sql.105%29.aspx