No, you just don't understand what they mean or how these things are tested. Actually go read DxO's various Q&A/FAQ pages explaining how they measure each aspect of a piece of hardware.
This page, specifically, explains the use case for ISO performance findings. For further reading, check
this one]http://www.dxomark.com/Reviews/Detailed-computation-of-DxOMark-Sensor-normalization]this one. For more information on how they test every aspect, not just ISO performance, simply read the rest of their explanations and FAQs.
If clicking a link seems like too much hard work and you want the lazy and dumbed-down answer, here you go:
High ISO performance is defined by the maximum ISO the sensor can go to while keeping a SNR of 30dB, a dynamic range of 9EV, and a colour depth of 18bits. The way you're most likely looking at your own files is just at the per-pixel noise, which is the same as only looking at the SNR. You're also unlikely to be factoring in how each sensor drifts from its stated ISO slightly differently. This is why pixel-peeping at home is not smart and tells you nothing, but it's also why lab tests don't matter. You don't see a difference; that's fine. That does not mean the lab testing is incorrect, but it also does not mean that it matters. If your pictures look good to you, fine.
