But you're implicitly agreeing with my point. Why are these other manufactured spheres important? Because they came from a different context! They're more likely to be made of different materials, and if so the sampling error of the small spheres being magnetized is eliminated. Having more sets of stones, taken from yet different contexts, would help even more. Exactly the same kinds of controls exist in radiometric dating.
I will agree that having such controls for radiometric dating is helpful. The fact remains, though, that no matter how many samples one examines, there is no way to eliminate the possibility of an unknown factors.
If one person conducts the shaker experiment by casting a set of spheres of different diameters from the same material, another conducts it by machining spheres of different diameters from the same material, another conducts it by casting spheres of different diameters from different alloys chosen to net the same weight, and another machines sheres of different diameters from such alloys, it's pretty unlikely that there would be anything about the smaller spheres--other than their size--which caused them to fall to the bottom. If all the spheres were machined from the same size blank, it would be plausible that for some reason machining a sphere longer would make it fall faster; if all the spheres were cast, it would be plausible that the smaller mold somehow concentrated the molecular structure in the sphere so as to make it fall faster. But by using different means of manufacture, such possibilities can be eliminated.