Thought - probably wrong, but I haven't seen it elsewhere.
Everyone says that dual wielding doesn't work, yet 2 weapon dorfs beat one weapon no shield dorfs.... if there is no advantage, this would not be the case.
It was stated earlier in the thread that two shields give two block rolls, so if the first fails, the second might still succeed, giving an overall higher chance to block... that post went on to say that no such advantage occurs with weapons, a you cannot hit with both weapons at once.
I would counter that argument with the fact that you cannot block with both shields at once either. Rather, you roll for one, and only if that roll fails do you get the attempt with the other.
If this were to hold true for weapons as well, then you would NOT get two strikes with two weapons, but you would have a second chance to hit if the first strike missed... netting an overall increase in your to-hit ratio.
Each actual strike would still do the same basic weapon damage, yet these strikes would happen more often than a single weapon would achieve, and therefor a higher overall damage output, just because each miss got a second chance. This could account for the discrepancy.
In such a case, a dorf that is well trained is likely to hit with the first weapon, and unlikely to ever get a roll with the second one; leaving no improvement... yet a poorly trained dorf will tend to miss more, and thus get more swings with his non-primary weapon, resulting in a net gain in damage output.
To be fair, the average dorf, even if poorly trained, would probably live longer, and therefor have a higher total damage output, if they were running a shield instead of the additional weapon... however, if true, this would make it situationaly advantageous.
For example, a bum-rush of newbies against an FB... the more cannon fodder you threw out, the less likely each one would be to be the target of any specific strike, making the shield less useful. Since their training is probably vacuous at best, a second weapon instead would give them a higher chance to hit and inflict damage on each swing.
In the standard case of bum-rush vs FB, obviously, the bum-rush dies... but that's often the point; kill off migrants, and do some preliminary damage vs the FB... in this case, this may be a method of increasing that damage.
It would also explain the variance in the data, and would make perfect sense, given the mechanics of shields.