Upgraded to firmware 0.9.1 and nothing changed in the voltage reading accuracy. I've 5 Shelly plus 1PM devices and only one of them is reading voltage values close to nominal (voltmeter reading), within +- 2 V. The other 4 Shelly devices are all reading way too low voltages (-10V to -20V) compared to nominal. I'm not sure about their current reading accuracy (or power) but I will check this in the next days.
It is frustrating not to have accurate power reading from these devices. What is the purpose of introducing this feature if not accurate! A Fritzdect 200 socket reads more accurate values.
Why not introducing a correction factor in the firmware (as is in the shelly 1 PM) or, even better, a user input calibration function (or curve)? Could such power correction be done in the script, does somebody know how to do this?