A blunt “real vs fake” split would be dishonest. Two distinctions govern every page, and three things are graded separately: plausibility, evidence grade, and tier.
The tier tracks one thing only: what the evidence record shows.
Whether a mechanism can work at all is a different question from whether it has been tested, and the atlas refuses to answer them in the same field. Physical implausibility is carried on a separate contradicts-established-science badge, which can sit at any tier.
That separation does real work in both directions. A device whose described mechanism contradicts physics but which no one has ever put through a controlled test is untested — and flagged. It does not get promoted into the tested-and-failed tier for a test that never happened. Conversely, a device can fail testing without contradicting any physics at all: the substrate is ordinary, the measurement is real, and what failed was the interpretation.
Collapsing the two would cost the atlas the sharpest thing it can say about a $3 “terahertz” stone. Under a merged label the stone reads as awaiting a trial. Under this one it reads as: nobody has tested it, and the mechanism as described cannot work. The second is both more accurate and more useful.
Even where a claim fails, the instrument and the raw measurement are usually real. What fails is the interpretation layered on top.
This is why tested — not confirmed is not a synonym for fraud. Electrodermal measurement is a real measurement; what failed blind testing was the claim that it identifies disease or intolerance (EAV/Voll, Vega). Corona discharge photographs a real physical phenomenon; the diagnostic reading of it failed (GDV/Kirlian). Transcutaneous electrostimulation is a real, cleared modality; what failed was the “adaptive” biofeedback doctrine layered onto it (SCENAR/DENAS). In each case the device keeps whatever it actually does, and a correct reinterpretation remains open.
The inverse holds too. Ordinary, working electronics can carry a claim that contradicts physics — the hardware being real is not evidence for the doctrine attached to it.
The grade is consensus on evidence. It is held distinct from physical plausibility and from the tier — which is why a device can be physically plausible yet grade low, or carry a real substrate yet sit at tested — not confirmed, or grade F while sitting in the middle tier because the refutation is physical rather than clinical.
If a vendor wants scientific credibility, these are the experiments — and most frequency-medicine marketing avoids exactly this: quantitative dosimetry.