Contrary to what you believe, there is no contradiction here. Implicit association test and stereotype threat (to name two large offenders) had reproducibility crisis. IQ and Big Five findings replicated a lot and basically have no reproducibility problem.