KM Editor not using the same text recognition algorithm for "evaluations"

I've spotted a problem. If I do this:

You will notice that the language specified is "Apple Text Recognition", but the phrase "(currently false)" is not determining truth/falsity by the "Apple Text Recognition" algorithm, but by the Tesseract "Languages - English" option.

I hope that's not too difficult to address.

P.S. I came to this conclusion because I'm getting too many mistakes in the evaluation, which suggests that it's using the wrong algorithm.