The claim is that when other people try to replicate the experiments from papers published in Nature and Science, the success rate is lower than when attempts are made to replicate papers published elsewhere.
(I have no knowledge about whether this claim is accurate.)
another study in _Science_, which examined a more general cross-section of psychology studies and reported an even lower rate of successful replication than the one linked earlier (39%): https://science.sciencemag.org/content/349/6251/aac4716.long
the authors of this study detail at least potential reason for the tendency of high-impact (i.e. widely cited) journals and studies to have lower reproducibility rates: major journals tend to prefer publishing "innovative" studies w/ results that have the potential to push the envelope and advance the state of the art. as a result of this position on the extreme cutting edge, though, those kinds of studies are less likely to have really solid results. at the same time, though, they're more likely to inspire attempts to respond to them in one way or another -- leading to more citations.
they note that this emphasis by high-profile journals results in a disincentive to put in the time to do the less glamorous work of reproducing other teams' studies, which is a potentially major issue for the trajectory of science as a whole.
It's an accurate claim, and has been borne out in research.
Science and Nature ... like publishing groundbreaking research, research with surprising outcomes, etc. These categories of findings are more likely a priori to be false, so it's not surprising even with a positive story they are less likely to replicate than a perfectly pedestrian study with an unsurprising finding.
