When you create the Escalation and add the escalation action that has the Is Breach box ticked, you are only really doing that for that specific escalation. You are allowed to have as many escalation measures as you want on a process, so you can create an Initial Response escalation and a Resolution escalation. Each of these will run between different statuses in the process and will have their own Is Breach setting. When the process creates it's escalation points, there will be two separate ones which are breach related. You can then test to see if either or both of these are triggered during the life of a process. What you must do is test whether the escalation action has been completed rather than look at the Is Breach flag that is part of the Incident object.
Thanks for you suggestion. So it seems we can only do that for reporting purpose.