There is also the concept of prompt and delayed neutrons. Prompt neutrons are products of the fission of the fuel, while delayed neutrons are products of other fission products and are released shortly after the fission of fuel.
Normally nuclear reactors run prompt-subcritical, delayed critical (meaning that each fission event causes on average less than one other fission event with prompt neutrons, but when you account for delayed neutrons then the average becomes one for one. The lag incurred by relying on delayed neutrons to push you up to critical allows the system to be controlled).
During both the SL-1 accident and the second Demon Core accident, for a very brief moment fission events were causing more than one additional fission events with prompt neutron products. In other words they both went prompt-critical, which is a type of supercritical. Thankfully SL-1 disassembled itself, the Demon Core rapidly heated and expanded, and neither were brought into a supercritical configuration quickly enough. This is the fundamental difference between what happened when the demon core went supercritical when Louis Slotin fucked up and when it went supercritical when the they detonated the bomb made with it over the Pacific.
"one of the design problems to overcome in constructing a bomb is to contract the fissile materials and achieve prompt criticality before the chain reaction has a chance to force the core to expand. A good bomb design must therefore win the race to a dense, prompt critical core before a less-powerful chain reaction (known as a fizzle) disassembles the core without allowing a significant amount of fuel to fission. "
http://en.wikipedia.org/wiki/Prompt_critical#Nuclear_weapons
The second demon core accident was a fizzle. It went supercritical but without involving a significant amount of the fuel.
The more pure you get your U-235 the more you can reduce the amount of time that you have to get it into a proper configuration. Pure enough to just drop together though? I find that incredibly unlikely. Keep in mind that even with incredibly pure U-235 your fuel is going to start to become critical when the two halves are still a few inches apart. Once that happens you have a very short period of time to close the gap and 9.8m/s^2 is pretty shoddy. The two halves are also quite massive, so being strong enough to slam them together by hand fast enough is not really reasonable either, they have too much inertia.
Maybe you could make a gun-type bomb using gravity if you dropped one half down a long evacuated tube... easier just to make it properly though. Of course refining pure enough U-235 that "simpler" gun designs become possible would be more difficult than actually just building the rest of the bomb correctly anyway.
I think Luis Alvarez was exaggerating "high yield" (you could certainly create a very damaging fizzle), or more likely "dropping", since constructing the gun is rather trivial with highly pure U-235.
My point is that there are many orders of magnitude difference between the relevant parameters of plutonium vs. uranium, so the fact that dropping plutonium cores against each other is insufficient to produce a large explosions tells us essentially nothing about that possibility for uranium.
The so called "Demon Core" went supercritical when Louis Slotin accidentally dropped a neutron reflector onto it (this was the second time out of three that it went critical) and although it killed him, there was no explosion per-se.
From wikipedia: "He quickly knocked the two halves apart, stopping the chain reaction and likely saving the lives of the other men in the laboratory, though it is now known that the heating of the core and shells stopped the criticality within milliseconds of its initiation."