Accuracy is a good choice of word. It doesn't come with the baggage of being a binary classification.
The terminology I jump to is "approximation error", where error is some quantified measurement of (in)accuracy. But using the word "error" might lead one to think of e.g. "having erred" or "being in error", which is unhelpful.
There's a quote from Box I like: "all models are wrong, some models are useful".
I guess this might be rephrased less snappily as "no model is completely accurate, but some models are useful".
Replace "model" with "theory", "belief" as desired.
That said, some models or theories fall into the category of being "not even wrong", i.e. to be so incoherent or unfalsifiable that it isn't even theoretically possible to measure how accurate they are.
Pauli: "Das ist nicht nur nicht richtig; es ist nicht einmal falsch!"
Better still would be to use "accuracy". But that presumes the audience is ready to understand the nature of the issue in a scientific and logical way. We want to get there, but we cannot start there.