Power measurement difference when connecting to other circuit breaker (behind the same RCD)

I've connected my Shelly EM as seen in below drawing. Color schematics are according to the Dutch NEN1010 (L = brown, N = blue, + = red, - = black). Both circuit breakers are connected to the same RCD, which is connected to 1 single phase.

I hooked it up to Home Assistant and starting examining the output via Grafana. During nighttime I noticed that the CT clam that was connected to the 7 PV panel grid, was returning around 50W. Which seems odd, as it was pitch dark outside. And the 3 PV panel grid was almost 0W. The Shelly EM was powered by the circuit breaker which also feeds the 3 PV panel grid:

So I decided to connect the Shelly EM to the circuit breaker where the 7 PV panel grid was connected to. Now the 7 PV panel grid returned/used 0W. But the 3 PV panel grid is returning/using about 19W:

During the daytime I also see quite a difference. When the Shelly EM is connected to the same circuit breaker as where the 7 PV panel grid is connected to. The power production is quite in line with that of the SolarEdge (API, 1 datapoint per 15 minutes):

But when I connect the Shelly EM to the circuit breaker where the 3 PV panel grid is connected to. The measured power production drops quite a bit below the line that is generated by the SolarEdge:

Why am I measuring different powers when connecting the Shelly EM to a different circuit breaker? Did I connect everything correctly?

Zitat

Welcome to the forum.

Can you please upload your pictures directly here in the forum?

Thanks

done!

This is what my fuse box looks like. First CT clamp is situated directly above the circuit breaker, in close proximity of other power lines. The second is a bit higher, and I tried to pull it away from the other power lines.

In below graph both CT clamps are on the same power line/circuit breaker, just to check if having the CT clamp into close proximity of other power lines has any effect on the measurement due to inductive energy flows.

The measured power difference is between 0 and 1 Watt. So this doesn't seem to be the issue.