I never understood this argument. I'm an Electrical Engineer, and I understand that there is an analog value that gets interpreted as a digital one or zero, but the implication that 'a weak one implies that the data was previously a zero' sort of assumes that the hard drive was only written once in the first place.
What happens if, in the course of writing legitimate data to the hard drive, the bit cell was set to zero, then set to one at a later date? Wouldn't the magical forensic tool get confused and come to the wrong conclusion that the 'weak one' was really a zero, when it fact it was just a 'weak one?' If you look at the analog value for one bit, you should be seeing evidence of the entire history of that bit, though in an indecipherable way.
I suppose it depends on how often the hard drive is being used, and how often it is being overwritten.