For the last one, either the measurements refer to different things (7cm = square perimeter and 3cm = circle radius is fine) and the researcher was just both lazy and bad at maths ...
... or it should really be classified as Safe for its ability to locally affect measurements, or at least heavily studied. Because if measurements are correct but incoherent ("Square shouldn't be in circle, wtf") and the reasoning is valid ("Size of square and circle imply an intersection there"), then either the theory is wrong and scientific research on the phenomenon should be strongly encouraged, which is most likely not the case otherwise the researcher wouldn't have felt it looked "wrong", or the measurements really are altered by the object and should be investigated the same.
Also, never underestimate measuring errors. "Faster than light neutrinos" anybody?
Rubber duck was aight tho.