You may be overthinking this problem. It’s true that the CT on your mains will accumulate the net power, while your meter as you say will only ratchet the import. So the question is what granularity of filtered sampling the net usage will reasonably represent your import?
Most solar installations flip between import and export very infrequently. They import Most of the time and export when the sun is bright. When you send the data to a service like PVoutput, the granularity is 5 minutes. So there may be some small inaccuracy in the 5 minute intervals where there was bothimport and export. If you upload to something like Emoncms or influxDB, you can get 5 second resolution.
If you want to directly measure the import at IoTaWatt’s sub-second resolution, there is a way that I think will work. If you put two CTs on the main and configure one to “allow negative values”, then one will measure (import - export) and the other will measure (import + export). Adding these together algebraicly we get:
Import - export + import + export = 2 x import.
Now if you divide that by 2 the result is import.
In IoTaWatt terms, let’s say those 2 CTs are named netmain and absmain. The output to get import would then be;
(netmain + absmain) / 2
If you are three-phase, you would need to do this only on the main with the solar feed if you have a single phase inverter (and add the other two mains for total import). If you have a three-phase inverter, you would need to do it for all three.
EDIT: btw/ even if there were a way to get just the import from a single CT, you would not be able to compute your usage from that. You would still need a “net” mains number to determine usage.