Mizuno's work: there are too many question marks, high risk of its being negative and if it is this would not help matters.
Takahashi: best of the hot H/? work. But evidence looks weak to me, and a negative would not be informative. There seems little agreement it is known how to make this thing work, which means google doing stuff with negative result is just big waste of effort.
Classic F&P / Mckubre - that would be best:
(1) This work has a better theoretical chance of being onto something
(2) The old experiments are still viewed by most here as "best published evidence"
(3) There has been a lot of work in this area and "how to do it best" is well understood.
(4) Most people here agree that with some care about materials replicability (at some measurable level) is assured.
FPHE is real => can get positive replicable results which will restart this area for serious research
FPHE is not real => negative results will be helpful too, preventing effort spent on stuff that will never work