I actually would think it's the other way around: ~10% of studies that sound like obvious nonsense and have p of ~0.049 are probably NOT going to replicate.