...I'm not sure how exactly you reached all of this, but for me, there's a whole host of issues here.
If you get a fighter from two times higher weight class - I think we can agree that's roughly two times stronger - than their opponents, that guy is going to wipe the floor with them even if there's three at once.
I don't agree that linear and square at the extreme ends of the spectrum. For a succession of duels, even if we assume you can apply this rule (this is going to be difficult because it is designed to work with a singular engagement, not a series of them), you will get a linear law only with massive assumptions - and I can't really call them simplifications because they don't work properly.
If you pit two fighters against each other, one of which is ten times stronger, it should mean, for simplicity and for the sake of the law, that this fighter deals ten times more damage and has the same amount of health. If we assume the damage is dealt continuously, and that it is proportional to remaining health, then the stronger fighter will leave the fight with slightly less than 95% health. In the end, you'll need about ten fighters in succession to bring the strong guy down to zero.
And that seems to work on paper, but I have two main issues with this. One, in reality damage is not dealt continuously. This works for armies because we can approximate them as singular entities with different properties than individual units, but you shouldn't make that simplification in a one-one fight, especially when you have non-linear relationships which arise because of non-linear damage resistances and speed differences.
Two, the assumption that damage is proportional to the remaining health is ridiculous for singular entities. For armies, this makes sense because it's roughly proportional to the number of fighters. For individuals, it just doesn't work like this, even before we include weapons and armor - and if you compound this with the above, you get a relation so drastically different that the law breaks down on a fundamental level.
Another thing, I'm not sure if you skipped over that detail or not, but I don't fully agree with the spear formation. It can be modeled locally with the square law, perhaps, but (with all of the above in mind) only on the assumption you have one-three lines of spearmen. Otherwise, the damage-to-health ratio gets broken and you can't use the square law with any accuracy. The square law works specifically because ALL individuals are engaged in combat at once, and this plainly does not happen in any mundane melee combat with a large number of fighters.
Finally, I can somewhat agree with the last line, but only on the assumption that damage is proportional to health, which, as said above, isn't really realistic. Realistically, the worst case scenario may work this way, but not because of the square law, but primarily because damage is not continuous.