Ovinomancer
No flips for you!
Okay, let's look at this. First, this only applies when the PC in question scores the final blow on the final monster for the combat and has attacks remaining. How often does this occur? In a 4 PC party where the PC supplies an equal amount of damage output, then it will happen, over an infinite period of time, in no more than 1/8 of all combats. This is because the minimum case for this to happen is a PC with 2 attacks who will equally share with the others the final blow, for 1/4 x 1/2 (the times where the final blow is leaves an attack). The party damage assumption isn't terribly realistic, but it's good for back of the envelope (spherical cows). So, DPR will be lessened by 1 attack 1/8 of all combats. Using the 3 round assumption, that's 6 attacks per round 7/8ths of the time and 5 attacks per round 1/8 of the time for an average of 5.875 attacks per combat, or 1.96 attacks per round, or an adjustment factor to DPR of 98%. The impact of this case, over time, is a 2% reduction in DPR prior to overkill consideration.Right, that's why I am saying the overkill effect on the last enemy in the encounter is equal for all PC's so long as DPR is equal. That's an important point in this discussion IMO and why I insist on looking at overkill effects in terms of an infinite trial of encounters instead of an infinite trail of enemy after enemy.
Set this figher's DPA to X. Overkill for this 2 attack fighter is (X)/2, and if we assume, as previously, that 1 killing blow occurs per encounters, that's a DPA adjustment of -(X)/12. DPR would be base 2*[X-(X/12)] or 2X - (X)/6. If we rearrange a bit, you get DPR = (11X)/6. Adjust this for the above, and it's .98 times that.
For the single attack fighter, DPA is doubled, but there's only one attack. If we keep X the same as above and just double it for this fighter, we see that the overkill effect is 2X-(2X)/6. Rearranging like above, it's 5X/3. This, right here, is the crux of the point I was making -- the impact of overkill on the single big attack vs multiple small attacks is slight, but it cuts against single big attacks. Even factoring in the loss of the attack using the above assumption set, the difference is still that the multiple attacks, same DPR outperforms the single big hit, ceteris paribus.