Hi
You need to review each graph for the duration of the test, overlay with additional graphs and locate the points where your CPU has exceeded its limit, or memory started to leak, or server errors were logged, or transactions started to fail rapidly.
it takes time and careful observation of the different graphs and you should know your application and your basic monitoring principles.
I recommend you go over the Analysis on-line documentation to learn about the capabilities of the tool and if you have specific questions after, we'd be happy to help.
Good luck,
Lior