Flesch reading-ease compresses sentence length and syllable density into one number, and it is useful precisely because it is crude. Some rough anchors from measured samples: a social media caption scores about 86, an email about 84, a blog post about 78, software documentation about 52, a research abstract about 24, and a law review article about 8.
The gap between the top and bottom of that range is not a quality gap. A law review article is not badly written; it is written for people who already hold the vocabulary. The score measures fit to audience, not merit.
Where it does bite is when the two are mismatched. Technical documentation sitting at grade 11 with 18 percent passive voice is usually documentation that could be grade 9 without losing precision, and the passive ratio is the faster thing to fix. Medical guidance in the same sample set reaches grade 7 at 8 percent passive, which shows that precision and plainness are not actually in tension.
The useful habit is to check the score against the reader you intend, not against a universal target.
A readability checker with these benchmarks is at belikenative.com.