The challenge with profiling is that you are getting some results but the interpretation of these results is extremely context dependent and possibly the issue is outside the context you are examining. It may be the tasks, it may be the glue between tasks and the graphs may not always tell the story. I find having some internal assumptions and being able to test them and building deeper mental model of the critical parts helps to flush out some culprits. Being able to at least eliminate some areas as "ok enough" can help to guide the search. It can be an iterative process where finding the right question and tool is more important than later looking at the data.