エマ - so I'm looking at what you reported. You are correct in what you recall (for おかわり): kanji > def was near 100%, and def > kana was high as well. You did not have multiples for this one, so it was just a transferal of mastery from the alternate version to the primary version. I, then, do not see any mastery data getting lost, at least on the backend. If you are not seeing those two mastery levels where you see the term, please shoot me some screenshots so I can look into it more. Might not be the mastery levels, but something else?
@Ace2 - Your schedules rolled over to the new day, but the kanji visibility issues should be fixed (and I'll roll out a patch for all users of the tool). It *might* have fixed the "not starting" issue as well, so let me know how that goes.
For my case: the versions of the words with kanji were duplicates, and the (previously unstudied) versions with only kana were chosen as the canonical form.
So, they don't have a mastery level at all for kanji -> meaning, because they don't have kanji. I'm planning to manually set their level for kana -> meaning, which wasn't a vector that I had been studying for words that contain kanji.
