It's two things:
1) i5-2500K and i7-2600K gives same performance in games because Hyperthreading is worthless for games, and
2) Margin of error between test runs. 2% is not statistically relevant anywhere, 5% difference comes still easy between two runs of same test in games that do not have pre-recorded benchmark demo.
You can see another funny thing with Hyperthreading when comparing i5-760 and i7-920, where the 760 is actually better. That's because games do not like HT at all, and can sometimes actually lose speed because of it even though both CPUs run at same speed.
Also one other thing worth extra attention in the chart is the performance of AMD x4 970 vs x6 1090T. Both of those are pulling roughly the same numbers, because again cores do not matter for games one bit. Because the x4 is much cheaper than x6, it's also better choice for a gaming computer, there are no ifs and buts about it.
edit:
Third thing worth noting is how high i3-2100 is actually in the chart, especially compared to AMD CPUs. Even though it has only two cores, games do not need more, and the raw clock speed combined with two generations newer technology makes it better performer than either x4 970 or x6 1090T, and it's also cheaper, driving the last nail in the coffin of AMDs being better in price or price/performance for gamers.