Presenting one story about studies that refute the phenomenon definitely doesn't turn it into "mumbo jumbo". The very same psychology you are calling "unscientific" is what you are using to debunk it.
There are other stories/studies out there, but I am not going to search for them. The story is using real-life data to debunk some theory.
Is because of what I have seen and experienced on mass transit, especially in NYC. No question that when someone is being accosted verbally, people turn their heads and don't get involved. In your link they studies that debunk the bystander effect are studying violent situations. Maybe that's a differentiator.
Maybe you are not understanding the underlying factor for the bystander effect claim. The larger the amount of people that witness something the less chance there is that one will get involved.
Let's take your subway example. Are you saying that if two people witness the behavior there is a greater chance one of the two will get involved than say if 200 people witnessed the event?