Well, there certainly are "corrections" that are fake. But there are others that are valid. It's really easy to do it wrong, though, so it's easier to get the right answer by discarding data instead of correcting it.

I'd like to see what would happen if someone spent five years gathering data without plotting it… seeing whether it was good data, making corrections where necessary (without forcing it to fit the trend, because they haven't yet seen the trend), discarding the bad data, and only then plotted it. But, sadly, nobody's going to do that, and I've got "better things" (heh, like there are better things) to do.