There are several reasons why the stories you read might not be conclusive science:
Striking results get more press, encouraging scientists to tweak their methodology until they can produce an experiment with big findings. One example is p-hacking, in which scientists collect a bunch of variables to try to find something that's statistically significant, even if the correlation isn't very strong. Read more about p-hacking from fivethirtyeight.com
Replication studies are almost never done. Although replication studies are science's best way to verify findings, they are rarely funded and under appreciated. Read more about science's replication crisis on Wikipedia.
The media often blows findings out of proportion. Reporters very rarely go through a study's methodology or explain the caveats because the original study is hard understand or locked behind a paywall.