I'll ask the question I've asked everyone who's said "all the factions are balanced" (or even 'ost OP'):
What's your explanation for the disparity in win rates? You can argue that the sample size for UKF is too small, but what about OST? This is from the recent World Cup Series, which had the same unit stats as the current version, and was played by top-level players.
First thing first:
You're right, you need a good population to be able to extrapolate data.
2nd there is an inherent flaw in this graph:
** How many games in total have been played? **
Other than that, you can then argue about the matches themselves. Were they anything else besides 1v1? The main thing people need to grasp is how fundamentally different a 1v1 game is to any other mode just by looking at the input parameters for each (let's say arbitrarily without any loss of generality) minute of gameplay. CoH2 1v1 is a different game altogether.
So yeah, the graph doesn't really show much, the percentages are there ad-hoc without any background information. If you want to build a good statistical analysis you need a large population, controlled system (tournaments are semi-controlled systems) and you need parameters which you would keep static. All in all, when you get such large disparities, you know something is off and the graph presented should be taken with a big fist of salt.
EDIT: Seen the link, 84 games. Not nearly enough to be able to draw any conclusions. Their statistical analysis is something that would be considered a bad representation/analysis in the scientific community. I mean, I don't blame them, it's a game that has countless parameters... Even though the thorough analysis they concluded is "much wow" effort, it's also waaaaay too much information and extrapolation. Way too much. The whole thing is convoluted, impossible to draw conclusions.
I'd say that the best way you would test the balance is if you had at least 1000 games. Only then can you proceed with an analysis that would not be inherently flawed.