I think this might be a case of throwing the baby out with the bath water. While there may be an academic argument to discredit the “Derived Reference” method of monitoring three-phase installations, I have seen no data to support it. To the contrary, IoTaWatt, where I believe the moniker was first used, enjoys tremendous success with this method of approaching polyphase measurement.
Site’s employing IoTaWatt for three-phase power measurement, and there are many, enjoy low-cost, ease of installation, and reasonable accuracy at both the mains and branch circuits. One prominent example of branch circuit integrity is the large number of both single and three-phase solar inverters that agree with the inverter’s counter with high accuracy.
I appreciate the argument that 1% integrity at the mains does not equate to 1% accuracy at the branch circuits, but by the same token, 1% accuracy at the mains does not indicate otherwise. One is not indicative of the other, but 1% consistently at the mains, across a broad base of installations, does suggest that a device has integrity beyond the mains. In fact, continuous IoTaWatt testing includes not only comparison to the mains meter (<1%) but measurement of several individual branch circuits that have loads at a variety of power-factors and are measured with various multiple CTs. Again, results are in the <1% range - daily, weekly, monthly.
I suggest that you look for inadequacies in the hardware or firmware of this device before declaring the method at fault. These arduino based monitors only have relatively slow 10 bit ADCs. You might want to try sampling at 75KHz as IoTaWatt does with 12 bit ADCs.
Obtaining 1% accuracy at any power level is not that difficult with suitable equipment. The differentiator is the firmware offering ease of installation, use, upgrade, and ability to deliver the data reliably.