It’s not a bug, it’s a feature 
During reviews, you are shown how many answers you have given correctly.
After reviews, you are shown how many items you got correct.
So if you review two items, and get the reading of one of them wrong, you will see 75% during reviews, because you’ve given a wrong answer to 1 out of 4 questions. Afterwards, you will see a score of 50% because you got 1 of 2 items wrong.
You might want different data displayed during reviews, and I agree they don’t make this difference at all clear. But you can’t bash it for being inaccurate.