This is a wrong understanding of p values. Imagine an urn with only fair 50/50 coins and taking out coins to test them. If you reject the coin being fair at 0.05<p then you will on average have 1/20 statistically significant results. All of your statistically significant results will be wrong.
Now imagine an urn that only has biased coins. Every time you reject the null hypothesis you are correct.
How many statistically significant results are actually true depend on the metaphorical urn that people draw hypothesis out of. If someone is very good at picking likely hypothesis to pan out their statistically significant results are more likely to be true.
Also if a field of study has many bad paths you could go down on then you would expect more statistically significant results to be actually wrong.