Once I had established the average power factor and voltage of my installation with which to configure the Owl (both of which are incredibly consistent from day to day), my Owl has proved extremely accurate for a whole installation over any reasonable period of time.
Since 5th April 2017, my supplier's meter has shown total Day usage of 18,533 kWh and total Night usage of 16,590 kWh. The corresponding figures from my Owl have been 18,552 and 16,622 kWh respectively - representing differences of about 0.10% and 0.19% respectively, which is plenty good enough for me. The same close correspondence exists of any time period from 24 hours upwards.
It makes no sense to say that they "tend to read high", since whether they read high, low or dead right depends entirely on what PF and voltage one programs into the Owl.