I have had my external CT's disconnected for a few days while moving them one notch closer to the incoming grid, they were placed after an incoming grid electronic voltage/current/frequency/phase fault detector which must draw something even if very little.
I have therefore, without external CT's also changed my setting from "no export to ct" to "no export to load". The ghost import is still the same 1KWh/day according to PEA, 0.3KWh according to Deye.
I have looked at the "purchasing power" graph in SolarMan and have disabled display of other graphs in order to get a good resolution. There is according to SolarMan a constant import between 0-20W but also a couple of quick but higher peaks (100-300W) during the day and they all seem to be related to Deye power tracking.
Example:
1. A high peak in consumption causes a high peak in import even though there is enough panel or battery power available to handle that consumption peak without importing from the grid.
2. A quick drop of panel power (a cloud arrives) causes a high peak in import even though consumption is low (1.5 - 2.5KW) and there is enough battery power available.
3. Rapid variations on the incoming grid causes a high peak in the import, slow variations in grid voltage doesn't do that but may be what is responsible for the up to 20W consumption.
The PEA meter did not move at all when I for weeks ran the inverter in off-grid mode and the inverter did then of course not record any import either.
So my conclusion is that the inverter resorts to import if it can (has grid connection) when it sees sudden changes in incoming grid, in consumption, and in production.
My OCD still want's to know where the other 0.7KWh is going though...