Recently, Gunn, Allison and Abbott (GAA) [http://arxiv.org/pdf/1402.2709v2.pdf] proposed a new scheme to utilize electromagnetic waves for eavesdropping on the Kirchhoff-law-Johnson-noise (KLJN) secure key distribution. We proved in a former paper [Fluct. Noise Lett. 13 (2014) 1450016] that GAAs mathematical model is unphysical. Here we analyze GAAs cracking scheme and show that, in the case of a loss-free cable, it provides less eavesdropping information than in the earlier (Bergou)-Scheuer-Yariv mean-square-based attack [Kish LB, Scheuer J, Phys. Lett. A 374:2140-2142 (2010)], while it offers no information in the case of a lossy cable. We also investigate GAAs claim to be experimentally capable of distinguishingusing statistics over a few correlation times onlythe distributions of two Gaussian noises with a relative variance difference of less than 10-8. Normally such distinctions would require hundreds of millions of correlations times to be observable. We identify several potential experimental artifacts as results of poor KLJN design, which can lead to GAAs assertions: deterministic currents due to spurious harmonic components caused by ground loops, DC offset, aliasing, non-Gaussian features including non-linearities and other non-idealities in generators, and the timederivative nature of GAAs scheme which tends to enhance all of these artifacts.