Perhaps you meant to day "science doesn't work by p-hacking and data falsification". The great thing about science is that it involves a process of building models and hypotheses, and then running experiments to produce results that support the models and hypotheses. To the extent that p-hacking and data falsification produce results that are not correct, longer term they will have a difficult time distorting our understanding of the underlying phenomena. Short-term, certainly they can be distracting, but we care about reproducibility because scientists are constantly trying to reproduce (possibly by building on older) results. Results that cannot be reproduced do not contribute to science.