Why is correcting a single character often a maddening experience?

When using my pen as input device I often run into situations where the character recognition is one character off. When that happens this beast often turns into a nightmare! Correcting a single wrongly recognized character is very often grotesquely impossible. :-(

I just had the case where if recognized an intended 'n' for a 'f'. I am already puzzled how and why the reco sometimes can be so far off. I mean: an 'n' and an 'f' have about as much similarity as a tree and a car. I then tried to correct (i.e. overwrite) that character. It first replaced it with an 'ö' (german Umlaut-O), then with a 'ß' (sharp s), then with a 't', an 'ü', an 'f' again and so forth. One variant more absurd than the next and absolutely no similarity with what I was inking and also far away from yielding any dictionary word! What's going off in this PoS?

If this reco-dance happens then the best is always to erase the ENTIRE word and rewrite it. That then typically works but it's so non-intuitive! If a recognized word is just off by one character it should also be possible to just correct that single character and one shouldn't have to retype the entire word!

Are these character reco-clowns at MS ever using this themselves?

And - yes - I already trained the character reco multiple times - in English and in German. That's the other annoying part, that the reco can't automatically switch or recognize different languages without *prior* switching. But that's another story...

Thanks - I did.
@Ztruker I like your signature line!