In transcription mode, my transcription of this sentence:
Elle était essoufflée et riait. Moi, j’étais sérieux, et je regardais ses prunelles noires à travers ses cils noirs.
was not accepted unless I used a curly apostrophe (’) rather than a straight apostrophe (’) in the word “j’étais”. My keyboard layout does not contain curly apostrophes, so it would be nice if either kind of apostrophe were accepted wherever either one occurs in the original.
Yesterday, I was on the verge of posting a bug report after running into the same problem with apostrophes in Vocabulary mode, also in French. You beat me to it @alanf_us 
In Vocabulary mode, the problem won’t come up in ordinary, single-word use, but I happen to use multi-word clozes in my personal collections.
I don’t think any keyboard has an easily accessible “curly apostrophe”/“single open quote” character. I can type it with a combination of compose+>+’, but I use an unusual keyboard layout in Linux. Otherwise, only programs using an “autocorrect” feature would transparently (and often annoyingly) change the character on the fly.
I think the sentences should be normalized by making sure only the ASCII apostrophe is used. It shouldn’t be that hard to do.
In a similar vein, normal spaces and nonbreaking spaces should be treated as interchangeable:
There are at least two approaches that can be taken:
(1) Allow each character of a set, whether that set is open quotes (straight or curly) or spaces (normal or nonbreaking) to match any other element in the set.
(2) Normalize the original sentences to only contain one member of the set (preferably the easiest one for most people to type) and require that the character typed in match that character exactly. [suggested by @mike-lima]
I actually favor (1), but even (2) would be an improvement over what we have now.
Note that one workaround until this is addressed is to copy the character from the prompt sentence into the sentence that you’re typing in.