And that last point highlights one of the major gripes I have with research of this kind. "Statistically significant" does not mean what you think it means. A lot of people see it as a license to draw conclusions from a set of numbers, which it isn't. Remember that even the standard p-test has an absurdly high cut-off of 5% chance that the data is pure randomness. And even then, even if some divine oracle rejects your null, you still have to deal with the fact that intrinsic error in research is still pretty much impossible to quantify. There's a reason that in particle physics they don't consider something truly proven unless there's a five-sigma confidence.
The fact that so many of these health studies show no hope of ever reaching a five-sigma confidence interval, only confidently proves one thing to me: whatever harmful or benevolent effect it is they're studying, its harm or benefit isn't going to be significant. Biological systems are just too unpredictable to make sweeping statements about the harm or benefit of many, many things. Even ionising radiation, the one thing we can all agree is super harmful in significant doses, becomes equally harmless in background-level doses.
So in the absence of reliability of single studies on cell phone radiation harm, I have to fall back on the next best thing: meta studies. And given that the overwhelming majority of research fails to establish a link, and the significance argument I made earlier, makes me very carefree and comfortable about the phone that sleeps right next to my head every day.