I have that same model, so i did quick test and i think you're right on both points - the glossiness map indeed looks like it's authored with 2.2 gamma and inverting it in photoshop does not make a proper roughness map out of it. I don't know why they took so odd decision, because it complicates things a lot. If glossiness would be authored with gamma 1.0, like it is in 99% cases, simple invert operation in photoshop would work correctly. Now however, your only option is to reverse the curve in Corona color correct, or to use output node.
P.S. is there specific reason why you can't or don't want to use glossiness and feel the need to convert it to roughness?