Looking at raw pk score isn't very accurate way to estimate the balance.
Well, it's accurate if you are trying to balance number of kills/deaths and only for this.
But a large part of pk is left out of this window. In general, str builds are much better at nailing kills, so if a stsf elf has the same number of kills/deaths as a fire/avalance/sword that usually means he won (killed or made his foe to run) many more fights that the giant.

While many don't consider making an opponent to flee as an achievement, I would argue about it, because it could mean that:
1) You lose your cabal item.
2) You failed to retrieve.
3) You escaped leaving your group to die.
4) etc.
For example, a necromancer usually has more kills than a paladin (and probably less loses too), still I don't necessary agree that paladins are weak.
If you will balance the number of kills for a necromancer and a paladin, then probably paladins will become too beastly.
Using this analogy: if a stsf elf and avalache giant have the same number of kills, that doesn't necessary means that they are somewhat balanced in power.

Also, raw pk score doesn't reflect many other factors like:
cabal, alignment, ganging, additinal powers, current set of foes, style, rp restirictions at killing, etc.
So when you look only at people with large number of wins and small number of losses you are at risk of gathering inofrmation only about a portion of players that have similarly favorable set of factors for their current character, leaving any other build/cabal/style combinations out of the review.

Not that I have a detailed methodology ready
So your short observation is still better than nothing. Just pointing out the limitations of it.