When computing cached hourly aggregates, the solardatm.find_datm_for_time_slot() function calls the solardatm.find_time_before() function to identify the starting row to use for reading difference aggregation purposes. That does not exlude rows that are missing accumulating property values, however. See NET-340 for a change impacting "live" reading difference calculations; this issue is to apply the same logic to the cached hourly aggregates.