This guideline is a bucket, not a diagnosis, and that is why the reply repeated itself — the reviewer often cannot tell you more because the rejection is a category rather than a finding.
What actually lands in it, in rough order of frequency:
Template or generator output. Apps built from a purchasable template are the thing the rule was written for, and they are detected by structure, not by looks. If any part of the project came from a template, that is the likely trigger even if you rewrote most of it.
Thin wrappers. An app whose function is mostly a website in a shell, or a single API call with a screen around it. "Small utility" is fine; "one screen calling one endpoint" often is not.
Category saturation with no differentiator you have stated. Not that others exist, but that nothing in your listing or your app says what is different. Reviewers are not going to work it out.
Account history. If the same account has published several similar small apps, the bar rises sharply. This is real and it is the one people never suspect.
So the question to answer honestly before resubmitting is which of those four you are. Rewriting the description and screenshots addresses only the third, which is why it changed nothing — if you are in the first, second or fourth bucket, no amount of listing work will move it.