Well most of the problems probably come from "Well this expirement was inconclusive but surely this mechanism is real and is the cause so I will just be a little selective in my data so it won't get dropped"

And after six iterations of this it just never pans out. Real replication is whether you can reliably build on top of a previous study.