When I first got a copy of DBA a comment by Phil Barker in the introduction struck me very strongly. He said that the previous, more complex and (apparently) more "realistic" and "detailed" rules (such as his own 6th edition) had given the player more information than a real general would have had during a battle. A general wouldn't know how exactly how many losses his side had surrered or inflicted. All he would know is whether a unit/body was winning a melee, fighting and holding its own but at a standstill, edging backwards looking over their shoulders or running for the hills.
So no more individual man at a time casualty recording, no complex reaction tests to be triggered by a wide variety of circumstances. Instead of troops defined by a micro-description of their training, exact gradations of armour and every weapon historically carried, define troops by what heir generals used them for. Have only three combat results - one side destroyed, one side pushed back and giving ground or both at a standstill. In other words an abstract system that nevertheless is capable of giving good results and (despite a few anomolies of its' own) very playable. It was a real eye-opener to find a set of rules where a game could play faster than a turn every 30 or so minutes. I hope no-one will be offended, but I see FOG as a descendant/relative of that approach to rules writing with added fine detail (amongst other differences, but it's the philosphy that matters here, not the details).
FOG PC is kind of both worlds - the combat system seems actually quite abstract, but the player is fed the level of post-combat casualty information the older-style rules presented. How surprising would the apparently wildly swinging combat results look if instead of percentage losses all we knew was the unit was steady, in some level of disorder or routed? Regard the disorders/fragmented states as indicating not simply that the formation is now askew but that individual men are starting to lose their nerve. Casualties aren't simply the killed and wounded, they include those who individually take to their heels and slope off even though the vast bulk of the troops are still steady.
I'd be particularly interested in the thoughts of the TT FOG people regarding this.
On a second point, "inferior" units defeating/causing high casualties on "superior" ones, and the fluctuations in losses as revealed by the game engine after combats, there are all kinds of historical examples of exactly that. As an example, my original wargaming interest was the 16/17th centuries, and there's considerable evidence as to what went on in pike v. pike fighting, and I can't see it being that different in principle in ancient times.
Say it's 1520 in Italy. Two blocks of pike, each 16+ ranks deep, files shoulder to shoulder, each man with his left shoulder towards the enemy, approach each other at a steady pace. The front 4 or 8 ranks level their pike and just before impact the whole body tries to compact itself front to rear so that each man is pushing the man in front of him. There are two reasons for this.
The first is that without 15 men shoving them the front rank might decide that seeing as the enemy have 8 pike heads levelled at every single one of them the job has lost it's appeal even with the extra pay - there are accounts of engagements where the entire front rank of both side's pikes was pretty much wiped out at contact so they were right to worry. The front ranks also were where most of the officers and senior NCOs were, because they were more experienced, often better armoured and they were meant to set an example; but as a result they were also horribly vulnerable. It was very important not to let the front ranks flinch because that meant you were likely to take heavy losses while not infliciting much damage yourself.
The second reason was that if neither side broke on impact there'd now be a mess at the head of the columns with pike shafts, dead, wounded and living men jammed next to each other so neither side would be able to use their weapons effectively - at which point the "push of pike" became literally that as each side tried to shove the other backwards. All very similar to the descriptions of hoplite warfare.
So - phalanx meets phalanx. At impact one side either mistimes it's final charge or momentarily flinches. It takes high casualties and inflicts few. The rear ranks however aren't affected anything like as much (being safe at the back), they put their shoulders to it and manage to halt the retreat of their own front rank and shove their opponents a few feet bckwards, allowing their side room to recover and use their pikes or reach for their swords. The situation is reversed, and so are the relative casualties. Up and down the battle front such events could easily be quite localised with each side winning some, losing some until eventually a gap appeared in one side or the other's phalanx. A satisfactory explanation? I don't know, but it certainly seems plausible to me. If FOG had the loser of a combat being physically pushed back as well as losing good order then maybe combats would be more understandable. And realistic

As for pikes v. legionnaries, if the legionnaries got off their volley of pila at the right moment then they pretty much had it made. Get the pila timing wrong and men holding (probably still drawing) swords either get piked then trampled or make for the rear.