My guess currently would be that you are seeing the effects of a marginal data block somewhere in the chain of medias (of which the memory card is least suspect as it has passed the polygraf alredy
). Actually, the problematic block might even fail quite consistently but then the effect of that could vary depending on what exact location a image file it ends up.
If you have the space you might let the copies that are created at various stages to to stick around. Then, when/if this happens next time you could go back in the chain of copies to locate the first copy without / with the problem and then take it from there: e.g. if the copy on the laptop is good, but the one on the external HD is not, then the HD would seem to be the one to blame.
I suppose H2testw could be applied to the other disks in the copy chain as well; it simply writes out a file with generated data on the disk in a location you can give, reads the file back and checks it gets the expected data back (and finally gets rid of the test file). The check part failing should mean having identified the culprit. Another thing you might do (for all the memory cards and hard disks involved would be "MyComputer - <disk X> - Properties - Tools - Check for errors" with the option "scan for bad sectors" checked in addition to "automatically fix ..." (takes a fair bit of time, as it goes through the entire disk). In case Windows finds "bad sectors" its marking them as such should mean that you image files do not end up using these (it is worth noting that the list of bad sectors is cleared with format).