Reviewers may prefer attentive staff, but the good reviewers (which exist) know that not everyone shares their preferences. That's the difference between a good review and a bad review: you can translate a good review into your own viewpoint and decide for yourself.
When I think of noisy one star reviews, I mostly think of the one star review for a microprocessor where the guy didn't use a heatsink and the motherboard melted. There are plenty of folk who come across well in writing but are monstrous in person, I'd never expect to trust their restaurant reviews, and I might not be able to identify them either.
I love reading Ebert's movie reviews. I often disagree with him, but he is a mostly fixed point. I can tell if I'm going to like something because, over time, I have developed an understanding of where his tastes and mine align and where they differ. This kind of thing is a lot harder to do online—it requires real work and attention.
In a way, we're disagreeing about where we want to have our statistical anomalies. Do we want the reliable reviewer caught on a bad day, or perhaps recognized by the staff? Or do we want the sea of one-time reviewers, a few of whom had a statistically unlikely bad or amazing experience?
The other problem is that not everyone is the same. For example I absolutely hate "attentive" staff in a restaurant (let me eat in peace and if I want something I'll ask for it), but want the check as quickly as possible on finishing while others like to slowly wind down. A regular reviewer probably won't mention either of these because that is normal to them.
The existence of multiple reviewers means one will be more like me and would mention these issues which would make me go somewhere else while others who enjoy that kind of experience would be more likely to come. This doesn't make anyone right or wrong, but does show how different reviewers help and why the lowest review scores have the most informative reviews.