If you want to argue that every single one of these replications was a mistake, you are arguing that the experimental method of science does not work.
The arguments are that the experiments are "noisy" and the signal is not significantly stronger than the noise and that in almost three (1989-2017) decades, this has not significantly improved. And for the better experiments which may have a stronger signal to noise ratio, replications are neither clear nor satisfactory. That is not MY argument. I am citing THE argument. And then, of course, there is the inability to scale it up. And then there is Shanahan who, despite Jed's strong negative opinion of his work, makes quite a bit of sense about the possibility of inherent, systematic, and consistent calorimeter calibration errors.
BTW, this is why the Ni-H folks were the most interesting. They offered high power out and low or very low power in and the potential of NO power in, removing any argument about legitimacy. Unfortunately, as we know, none of that expectation proved out and the largest players were actually crooks. But everyone know all that, no?