There are 2 cohorts that should show the same results, but do not. It appears that the coded observation cohort includes voided patients. (It correctly does not include voided observations.) When using a sql query cohort, it can be written to avoid inclusion of the voided patient/obs.
When writing a composition cohort (to show the additional patients), not difference appears between the coded obs and sql query. Perhaps the composition query does test for voided patients?