Well, here's the kind of thing I mean:
Lets say you want to test the phrase Ivory Soap floats.
So you take that as a hypothesis and design an experiment. You get a tank and fill it with water and put the bar of Ivory Soap in the water. Then, let's say you do this experiment 1000 times. It floats for 998 times, but for two times it does not float.
Now, wouldn't the overwhelming evidence be that Ivory Soap floats? And wouldn't that also make it a theory?
Because, if we take the philosophy that in order to see if something is wrong with the experiment, that is to say there is any kind of mistake what so ever, then somebody can point to those two times that Ivory Soap didn't float that it would call into question the entire experiment so that people can say things like "it's impossible for Ivory Soap top float".
That's what is confusing me about what you mean by falsification.