In my direct experience with this phenomenon, it is due to the relationship between tail length, frequency response, and minimum phase processing. Most people think it has to do with algorithmic tools or outboard factors such as preamps, eq's, converters, etc, but it does not. I'll give you a quick example:
For the Room IR's I'm working on right now, I have them at 500 ms for the full resolution files. When I do my manual time/phase alignment to the OwnHammer standard, a normalized deconvolved file comes out around the same as everything else. Add minimum phase processing to the same process, and the initial spike of the impulse raises greatly in relationship to the rest of the data and as a result of the time at which is taken into consideration for the normalization math, the output is significantly lower. How to work around this? Shorten the tail length. If I take this same minimum phase IR file and prep it for Axe-Fx I & II format by truncating the tail and then normalizing it again, it's fine, as the duration of the file and the initial spike is all similar to the near field captures. This may not be a viable option for you, as it wouldn't be for me if I depended on minimum phase processing for my full resolution .wav IR's. One of the many reasons I'm against minimum phase processing as I am for it.
The converse I'm sure everyone else has experienced with files that have a lot of bass, they can be way louder than normal files. I wrestled with my upcoming back of cab files for a couple days trying to find a placement and eq setup that gave the sound, character, and response I wanted but that wouldn't blow your woofers switching to it from a typical near field. I think as they originally were, with them normalized the peak value was between 12 to 15 dB louder!
So again, it's not the equipment you're using, physical or intangible, it's the math behind the processes involved and how they interact with your collected data. This will be something you will either fight with, accept, or be forced to find some kind of compromise.
Hope this helps a little.