League-of-legends – ny difference between Cleanse and a Mecurial Scimitar

As I mentioned in the comments, you're looking at the wrong aspect of armor. You shouldn't be looking at the percentage it reduces, but rather the amount of damage it actually prevents as a measure of how much longer that armor lets your survive.

Let's consider an example for a champion with 0 armor. Say he has 1000 health, just 'cause that's a nice round number. Now our champion gets caught in Garen's judgement, and starts taking 100 physical damage each second (at one hit per second). Clearly, our protagonist can take ten hits before dying.

But now, imagine our champion has the 50 armor you mention, meaning the incoming physical damage is reduced. Now (still with 1000 health) it takes 15 seconds for that bastard Garen to claim our champion's life. 50 armor extended our life by 5 seconds. (100 * (2/3) = 66.6 DPS; 1000 / 66.66... = 15 hits)

So our champion respawns, wises up, and buys 50 more armor. Now with 100 armor, we're sitting at a flat 50% damage reduction. Again we run into Garen, who's still dealing 100 base damage per second. It now takes him twice as long to kill us -- 20 seconds. Once again the armor has extended our life, again by 5 seconds. (100 * .5 = 50 DPS; 1000 / 50 = 20 hits)

Back in base, we buy another 50 armor. We're now at 150 armor, and 60% damage reduction. Enter Garen -- spinspinspinspinspinspin and it takes 25 seconds before our champion dies. Again, adding 50 armor extended our life by 5 seconds. (100 * .4 = 40 DPS; 1000 / 40 = 25 hits)

So even while the amount of damage reduced per point of armor diminishes (as you observed), the amount of time armor extends your life (or effective health) remains distinctly linear.

Okay, from my experiments, it goes:

Armor/MR reductions <--> Percentage Reductions
(They stack multiplicatively, so order is irrelevant)

Flat Reductions

This makes sense, since if flat reduction was the first thing applied, after armor, it may have failed to provide a benefit at all (since, for instance, the difference between 50 * 27% and 48 * 27% is minute at best. 50 and 27% being the base attack damage of Ezreal and the armor of Amumu, my two test characters.)

Shields act as health, and damage done to shields is likewise reduced by armor / magic resist. (Meaning shields on tanks can generally absorb more punishment than shields on squishies)