Biased reading by voltage increase?

I have a 41kWp solar installation on a building with EV charging consumption.

Before installing a few OPENEVSE charge stations, I am monitoring production and consumption with a emonpi.
I just measure single phased , both in production (one of the three phases of the inverter, which are balanced) and consumption (same phase that I measure on the inverter, to use the same AC/AC reference).

Now, the issue: I see an unexplained consumption phenomenon, which is an image of the solar production (see picture attached). I wonder if this is due to the voltage increase that a solar inverter creates, when on a grid with small loads.

Anyone has sse this issue before?



p.s. first i thought there was some stray magnetism from production which could create a false reading, but now the two CTs are well separated:

On the 2 occasions I’ve seen this I’ve found a CT was not fully snapped shut or that the CT was being pressed on by a neighboring cable, in both cases there was a slight air gap in the split core causing a difference in readings eg the PV is read as 1kW by the PV CT and also as 1.2kW by the Grid CT if exporting (if wired that way, I cannot tell from your pics).

So assuming the 2 CT readings are used directly in your graph and there is no additional processing to throw things out, I would check for air gaps and/or cracks in the CT split cores.