I would recommend you this userscript. If it really bothers you and interferes with your studies, it’s better to look at numbers after the review session ended.
As to why two percent values are different. It’s because they are different values, wild huh. The top right corner is a ratio of failed questions to all questions asked, so the meaning and reading count as two different questions. The number on the post-review screen is a ratio of downgraded items to all asked items. And 80% is totally all-right.