The total reviews will be less than ten percent of the installs. The negative reviews are a fraction of that percentage.
As such you cannot observe via reviews what the majority of the users think of the app.
Then, by the definition you have just given, app reviews aren't empirical.
Sorry, but none of your badly formed, hand-wavy reasoning has proven that wrong. Also, nobody is arguing that "the majority" think something - I'm arguing against your completely anecdotal and un-evidenced claim that it "rarely" happens.
What evidence do you have that it rarely happens? None that I can see so far...
It's like surveying a small percentage of an electorate by asking them who they'll vote for for and deciding the result based on the survey.
Anyway, once again... where is your evidence that this rarely happens?
My evidence can drawn from the fact that ten percent of people write reviews. A fraction of that are negative. If an occurrence happens say five out of a hundred times,that's a rare occurrence.