There is even more to that.
Sometimes, conventional bombs whose fuse malfunctioned could be found more or less in one piece. A skilled, lucky unexploded ordnance specialist might disarm and secure them, at which point the bomb could be studied and/or the payload used against the producer.
But some of those unexploded conventional bombs had been unloaded at low altitude, and thus had not reached terminal velocity. Some had a parachute, further slowing them. Virtually all of them were not intended for air burst, but to hit the ground and then explode; some were intended to dig into depth (say, though a tall building's upper floors, or right into the ground, or into a RC submarine pen top, or into a battleship's armored deck). This meant they dedicated a sizable proportion of their total weight to a thick casing, rather than to pure explosive payload. Such a casing equates to them being armored, which increased their chances of not breaking up upon impact. And finally, virtually all of them had one fuse.
Fat Man, OTOH, would have reached terminal velocity when hitting the ground. It was intended for air burst, and though it did have the equivalent of armor (for reasons linked with nuclear physics, and also as a precaution against any stray AA fire), it was not that thick. It had multiple, redundant fusing devices. And finally, those fusing devices activated no less than 32 detonators.
So the sequence needed for the Japanese to retrieve something is:
- the redundant multiple fuses do not work;
- the bomb does not break up on impact;
- the impact does not accidentally activate one of the 32 detonators;
- the Japanese bomb disposal crew also does not.
My own assessment of the likelihood of this sequence is under 0.1%. Definitely, a dirty-bomb "fizzle" is the more likely result of a malfunction.