Having recently replaced my solar system, I now use the tesla inverter.
Looking at the daily solar production from both measurement over the last week, I noticed that the Tesla inverter is over-reporting production compared to Sense.
The Tesla production is internally monitored by a Neurio and CT inside the inverter, but only on one phase. while the Sense is inside the Electrical panel monitoring both phases.
I think the Tesla (similar to the overconfident range specs for Tesla cars 
), is slightly cheating here. It looks like they are including (probably with software tweaks!) the power used by the inverter in the “production”, even though that’s just the cost of doing its business. The real production report should be the net amount available to the user, not the raw solar production. So basically the inverter itself consumes ~ 1.2kWh over about 11 hours. Of course the Sense reads negative all night, which also lowers the daily tally, but that is just about 6W most of the night time. (The quoted CEC efficiency of the inverter is 98%). Yes, the inverter is liquid cooled and gets quite hot, so that power has to come from somewhere.
As mentioned, the sense monitors both phases while the Tesla monitors only one and “calculates” the other (how?). See here.
I do get slightly different powers on the two branches, but that might partially be due to the voltage difference.
I wonder if my analysis is correct or if there is an entirely different explanation.
Here’s a graph and linear fit of the reported data.
