Good study design trumps replication. If you construct a shitty misleading study and everybody jumps on the hype train to replicate you've netted negative on the replication axis.