it happened once that even though translated text existed in the terminology database it has not been recognized by QA check.
After checking the source of terminology, as well as the translated text it turned out they're not the same characters. By looking at this site one can identify unicode characters : http://www.babelstone.co.uk/Unicode/whatisit.html, here's was identified in this very case :
The term in the terminology database corresponded to the following characters :
so we're actually seeing the letters "O" and "C" but they are part of the Cyrillic alphabet, whereas in the translation, we find these results :
which are the classic "O" and "C" characters of Latin alphabet. It therefore explains the fact that matches are considered different and that the QA module identifies this as a misspelled word. While they're visually similar, they're not the same characters.