30% is the error rate for measuring radiation AT the detector. In doing something like comparing pieces of fish, with all the factors mentioned above, I'd expect monumentally greater error rates, of several thousand percentages... rendering the device useless.

There is plenty of evidence that the stress created by a device like this, by an obsession with infinitesimal amounts of background radiation, does far more harm. I see no net health benefit from such products.

Fyi, a 25% error rate detector can be build using a soup can and some string. (Not a joke. This thing actually works.)

https://en.wikipedia.org/wiki/Kearny_fallout_meter