I have always seen the tier system not as a ranking system, but as a classification system. It is not as much the relative power of the classes, but can be seen as the relative reactivity of the classes. It is more akin to the periodic table of elements than a list of tournament rankings. for the purposes of this analogy we can consider an encounter to be a known set of random atoms. (NOTE: For the purposes of analogy, most of the chemistry is simplified to basic levels)

Lets call the tier 1 classes the alkali metals, we'll use lithium. When you take any set of random atoms and set it as an encounter for lithium, the results can be difficult to predict. The lithium can react with nearly anything, making for a very interesting encounter but a very quick one. The numbers of options that the lithium has is immense, just like a standard tier 1 class.

Now lets look at a encounter between another known set of atoms and an alkaline earth metal, magnesium in this case. Magnesium has fuller valence shell, so it is not quite as able to react to nearly everything. It will still react with something, but it is slightly more predictable than lithium. The number of potential reactions is still staggering given a complex enough "encounter', but it will always pale in comparison to those available to lithium. This is equivalent to a tier 2 class.

Let us skip a few now and look at the halogen family, namely chlorine. With a nearly full valence shell the ability of chlorine to react is extremely limited compared to magnesium and nearly incomparable to lithium. In a reaction with a given "encounter" the possible interactions of the chlorine are easily predictable. Here fall our lowest tier classes.

Actually the preceding statement is a lie. The lowest tier of classes in our metaphor is the noble gasses. Given nearly any possible random encounter, a noble gas will do approximately nothing. With a full valence shell, there are nearly no possible reactions for neon to be a part of. It will have almost no impact on the overall situation. This is the true lowest class; the unoptimized monk, the poorly played truenamer.

In closing, I believe that it is in fact a fallacy to see the tier system as a measure of power rankings in the classical sense. The tier system is not a measure of who has the most plusses in an individual category. The system is using power in the same way that I have been using reactivity. The system measures how many options a character has. The power comes more as a byproduct. In a given set of moves, some are more powerful as a matter of statistics. In a big enough group, the outliers can be very extreme. The power of tier 1 comes from these outliers, which is by definition only existent due to their large availability of the moves.