I think that’s an absolute certainty, given that you have no on-site generation.
There has to be a rational explanation, but nobody - me included - seems to be able to find it. Something has clearly changed, either you’re discounting it or you’re simply not aware of it, but between us I think we’ve covered all the possibilities that we can think of.
Let’s go back to basics. The direction of power flow is given by the relative phase of current and voltage. (The full explanation, if you need it, is in ‘Learn’.) You can change that by reversing the c.t. on its cable, by moving it from line to neutral (while still facing the same way) or by swapping the leads inside the jack plug, or by doing similar things with the a.c. adapter: swapping line and neutral on the mains side or swapping the wires on the low voltage side. Once the voltage and current get into the emonPi’s front end, the power comes out of the maths as positive or negative depending on the phase relationship, so from then on it’s a case of a minus sign in the wrong place. So if you have a negative calibration constant in emonHub, or a minus sign in emonCMS’s processing, either of those will give a negative power too.
It has to be one of these things - somewhere. Or an odd number of them. An even number will cancel.
It can’t be something the electricity supplier has done upstream of you.