Dear Mary,
You are right, of course, that even with the combination of two normal distributions a "gain" part of the wall loss distribution inherently remains (I told about the isolated inspection), but "scaled" by the second distribution. Btw, consulting the Recommendation ITU-R P.2040-1 Annex 2 Section 2.1 confirms this attribute of the normal distribution of a wall loss.
Regarding your dRSS distribution, with a standard deviation of 34.5 dB at a distance of up to 40 m any path loss model will turn to a path gain model, even though with a low probability. Concerning the figure of 34.5 dB I guess that you are talking about an indoor-indoor propagation (as indicated on a recent email). However, I'm quite surprised about this large number, in particular as the respective Recommendation ITU-R P.1238-8 considers just up to 17 dB at the 5 GHz range. Which model do you apply?
Best Regards,
Karl