How frequently do external evaluators find areas of improvement that are not obvious to internal evaluators?