As a newcomer to profiling you should start by simply looking for methods that have a long runtimes and/or are invoked many times during typical usage patterns/where the bottlenecks occur.
I am not sure how the Eclipse integration with JProfiler works, since I primarily use NetBeans. However, in NetBeans there is a 'Snapshot' view that shows a hierarchy of method invocations with runtimes that sum up to 100%. I look for the parts of the hierarchy that take up a (relatively) large % of the total time. From there you have to think about what those methods are doing, and what could be causing them to be slow.
For example: I noticed that a method that was called frequently was overall taking way too much time to complete, and was a serious bottleneck. Long story short, it turns out the code was checking to see if an item was present in a collection using the .contains() method, and the collection was a Linked List. The reason this is a problem is because Linked Lists have time complexity of O(n) for functions like .contains(). The fix in this case was quite simple as I was able to replace the Linked List with a Hash Set, which performs .contains() much faster, in O(1) time.
