Currently all aggregation functions timestamp at the beginning of their applicable periods. This is at odds with the behaviour of most dataloggers, and makes the use of aggregated variables in graphs alongside logged values, at best, confusing for the end client.