To begin with, you must ask whether the devices that you are comparing exhibit the same measurement accuracy. Thus, when comparing measuring devices with different measured value acquisition (in this case, inverters and EVU counters), the results differ from one another.
In general, the measurement uncertainty of a measuring instrument relates to the respective measuring range end value.
This can be made more clear with the help of an example: For the Sunny Boy 2500 inverter, a measuring range end value of 3000 W, with a maximum error range of ±3%. This means that there is normally only a measurement accuracy of ±90 W. Smaller power measurements are also detected with this deviation. Here, the relative measurement error is considerably greater (in the case of 900 W instantaneous power, a ±90 W deviation corresponds to a relative error of ±10%).
In contrast to an electric utility company counter, a calibrated measuring tool with a measurement accuracy of max. ±2%, our inverters are primarily designed for feeding electrical energy into the grid in as effective and cost-efficient a manner as possible. The displaying of operative values is more of a secondary function and is therefore intended for rough qualitative judgments and quantitative estimates regarding the operation of the device.
Additionally: No PV inverter on the market uses measured value acquisition that functions with the precision of a calibrated feed-in meter.