Most probably a measurement/calibration error. I would expect the smart shunt to be better calibrated than Multiplus’ internal one. In any case, 0.5% of a difference is an extremely low value for anything but rocket science or nuclear plant management, imo.
Loads around 4A on the out put are a bit out of range or hard to read as it is done with a hall effect sensor. Power factor can also affect it at higher loads. (Which is why there is a low limit in what you can also set as the input current limit)
Anyone who is serious about inverting though would be interested in the battery readings mainly as that is the base of the system.