Scope and limits

A MODELLED result from a 2026 arXiv preprint whose peer-review status is not established; applies only to the specific 1 GW, 2.45/5.8 GHz, mid-latitude, vertical-incidence, quiet-ionosphere case modelled. The paper flags oblique incidence, geomagnetic anisotropy, and disturbed/storm-time ionospheres as future work not yet done. A POSSIBLE SOURCE-SIDE DEFECT AGAINST THE 3585 K FIGURE, raised by VER-1094 and NOT resolved. SRC-105 states at printed p.9 that its neural surrogate is trained on |E| from 1e-2 to 500 V/m and at printed p.10 that inputs are "clipped to the training ranges", yet its Figure 13(a) at printed p.21 shows the 5.8 GHz peak field at roughly 532 to 614 V/m along the ENTIRE path - above the stated ceiling everywhere - while the 2.45 GHz case sits at about 234 to 248 V/m and is inside range. If the clipping was applied as described, the 5.8 GHz maximum dTe and maximum Qh are biased LOW, and the paper's limitations paragraph does not mention it. Not asserted as an error: Table 1 shows separate surrogates per frequency, so the range may have been widened without being restated. What is established is an unreconciled discrepancy between the stated method and the plotted result, and it bears on any figure this claim takes from that column.