This claim is very controversial to me.
For instance, in the last session of Torchbearer that I GMed, the PCs piled scrap metal onto a magically-conjured floating disc. Presumably this falls under the metaphor of the players "interacting with" the scrap metal. But the scrap metal had no statistical description.
The PCs also drank water (no statistical description required), lit a fire (no statistical description required), and journeyed from a ruined tower to a forgotten temple complex (no statistical description required). Their journey was hindered by blustery winds, whose relevant statistical characterisation was +1 toll. The PC with a cloak was able to reduce the toll for the journey, based not on anything statistical but on common sense reasoning about the fiction.
In general, the way we assign physical properties to things in the fiction is by describing them. And those descriptions then provide the context for the players to declare actions for their PCs. Only in some contexts of action resolution do we need game mechanical statistics, and those statistics are generally relative to a particular game play context (like penalty to the toll from a journey).
Does it? Have you done the maths for every instance of minionisation?
In 4e D&D, a 1st level standard NPC has AC around 15, hp around 30, around +5 to hit for around 9 damage. A 9th level minion NPC has AC around 23, and around +13 to hit for around 9 damage. Whether that minion is easier or harder to defeat is not obvious at all. For a 9th level ranger with +13 to hit (+5 stat, +4 level, +2 enhancement,+2 proficiency) and AC 24 (Hide +2, +4 DEX, +4 level, +1 sundry), it will take two attacks on average to hit (and kill) the minion. The minion may get in one attack, with an expected damage of 4 to 5.
Assuming the ranger does d10+2 on a hit (weapon +2 enhancement), it will take around three hits to kill the 1st level standard NPC: 16.5 (weapon dice) +6 (3 hits) +7 (two lots of Hunter's Quarry). In that time, the NPC will get in (say) two attacks, with an expected damage of about 1 per attack, or 2 overall.
Of course other examples will come out a bit differently, as the possible variety is pretty significant, and the level of the PC relative to the NPCs will affect the maths quite a bit. But I think this illustrates that the general effect of minionisation is not to make an enemy easier to defeat, but to make the maths of the game more workable. (Ie the exact opposite of your "kludge" claim.)
Actually, in 4e D&D how tough a creature is and how hard they are to kill is represented, statistically, by the combination of AC (or other defence) and hp. And these are adjusted relative to the power of the PCs, in order to make the maths work smoothly. If all you know of a 4e stat block is the number of hp, but not the level of the creature (and hence its AC and attack bonus), you don't know how hard that thing is to kill, how tough it is.
This is a very obvious feature of 4e D&D.
I'm not sure if you ever played 4e D&D, but it sounds like maybe you missed the rules for level scaling of to hit, AC, damage etc. You seem to be assuming that "minionisation" leaves the stat block unchanged but for the creature hit points. But that's not the case (as my example just above illustrates).
Mechanics do matter. Changing the stats of the creature/NPC, to reflect the relative power of it and the PC, so as to make the maths of the game work smoothly, does affect the feel. In my experience, that is a good effect. Just as one instance, it tends to reduce spike damage from weaker foes - the scaling of minion attack bonuses together with the flattening of their damage numbers makes them more "even" in their contribution to the combat. (Whereas using a lower level standard means they mostly whiff, but very occasionally hit or crit for less "flat" damage.)
This helps support the Conan-vs-pack-of-were-hyenas feel.
It's not confused. Nor confusing, at least in my experience. It is one solution to the problem of using a relatively uniform mechanical resolution framework, and a reasonably linearly scaling PC build framework, to handle both a hardy young squire fighting off a bandit or two and a demi-god fighting off a flight of Vrock demons. I don't know of any other version of D&D that handles these things particularly well.
Assuming the players understand the rules of the game, they know that the Ogre's stats reflect its power relative to their PCs. Being immersed in the fiction (presumably, if we're successfully playing a RPG) they will appreciate that the Ogre would kill commoner's willy-nilly.
The same reasoning applies to the Ogre who steps on the cat's tail.