With only a little bit of hyperbole, that's because the question "is Coffee Good for you or Bad for you?" isn't scientific. At best, science can try to answer

With only a little bit of hyperbole, that's because the question "is Coffee Good for you or Bad for you?" isn't scientific. At best, science can try to answer the question "Do people, that are very similar to you, perform better on some metric when they drink coffee or don't drink coffee?" Especially when it comes to food, there are so many variations on what "good or bad" looks like, and what "you" looks like, that we should expect different studies to come to different conclusions they look at different populations, and measure different things. Even when we try to measure the same thing, on similar populations, it can be really hard to isolate all the important factors. Even then we expect different results. The problem you are talking about is not a science issue, at least not directly. It's a Popular Science issue, or at a stretch a Science Communication issue. Everybody loves a powerful model - a simple description that explains a lot - but those are rare and far between in science. In attempting to reach for new models, such as modelling the effect of coffee on human lives, there is a tendency to take scientific results and reduce them so that they look like a powerful model. Coffee is Good! Coffee is Bad! This process is not science! At best this may spur researchers to test if the model is correct, but more often it's just fuel for press releases, popular science journalists, and social media battles.