Why is there a discrepancy between the Top SQL report in Performance Analysis and the Oracle awr reports? PA shows a query executed ~8000 times, while looking at approximately the same window (even less) with awr reports, the actual # of executions is ~47000 times.
The reason for the discrepancy is due to the way we collect data. We only sample the data at specific intervals, therefore we do not catch sql statements that execute between sample intervals. The executions we do catch just happen to be the ones running at the time we sample the sga. It is helpful to remember that PA is not an auditing tool, but a performance analyzer. Therefore with regard to sql executions, PA is primarily concerned with those sqls which run for extended periods of time.