Seems to have worked pretty well for a beginner's LoRA, though there's some flaws. A few pieces of advice

1: I would not use a common, English, word like “crescent” as the trigger. Crescent symbols show up because of it.
2: “quad tails” is a real tag with lots of uses and should have been used
3: It looks like “kimono” was only used once, but you describe it as the main outfit.
4: When working with comics and manga data sets, blank out speech bubbles in training data (cover them up with white) and use the tags “speech bubble, blank speech bubble”. It works way better than leaving them in unaltered or trying to work around them.