Say I'm looking for a mapping application and one doesn't even have data for where I live, or some other attribute that would cause me not to consider it in 6 months when I look for mapping apps again.
I do the same for restaurant reviews, I might give one 1 star even though it serves palatable food, just not the kind I like.
By doing so the recommendation algorithms that harvest my feedback have access to both positive and negative feedback pertinent to me to give me better future results.
That's how those kinds of scoring systems work best, not by averaging everyone's scores and pretending we're all robots that (dis)like the same things.
What you're saying is true, which is why reviews are context-dependent.
E.g. this user lives in Germany, so his review has more weight for Germans. This user likes Indian food so his review has more weight for people that also like Indian food.
Well that would be the ideal. But then you need the ranking process to know something about you, at which point you can't really work with real anonymity. Maybe Google could hide the real names and profiles of people that want to remain anonymous, but use their profile in the ranking they are doing.
Sometimes the provider might design their star rating system purely as a consumption experience, not as a recommendation signal. Not every product has the data or reasons to use star ratings as recommendation inputs, and I'd expect that most people look at star ratings next to an item and use that as a quick gauge of whether it's good or not.
Assuming the amount of traffic (drive-by glances) is considerable, compared to the number of times you personally get recommended something. Should users submit star ratings for their own purposes, rather than the greater good? I think there are several arguments for "no", but the murkiness is the real catch. If people think their stars are being used for Netflix-style "We think you would rate this 4 stars", but the only actual use of them is to show other people how the product ranks on an arbitrary scale, then both sides lose.