I've always been surprised in my relatively short membership here to see how much mods are discussed but how little their actual validity is discussed. It's good to see some consideration being given to it here.
Could I throw in another line of thought? That is, perhaps we need to consider what end result we're looking for here. Do people buy modified versions of gear so that they can feel confident that what they are using is the best? Is it so that when they play their recordings to other people, the listeners will be more impressed by the recording to noticable degree than they would otherwise? Is it because they prefer the sound of the modified unit, regardless of whether it's actually more or less high fidelity?
I think it's necessary to specify the objective before getting too deep into the pros and cons and considerations of measurement vs subjective evaluation. If you're simply going for what you personally like, then of course the measurements might be irrelevant.
My take on the whole thing is that most of the gear discussed here reaches levels of audio accuracy such that making any improvement suffers from the law of diminishing returns. The chances of clearly enhancing the record/play accuracy (in other words, we're not talking about taste but about correctness, probably measurable) of the equipment by a mod are slim. There are examples of obviously poor designs out there - for instance the mic input of a Zoom H2 is obviously and inexplicably noisy - and if somebody came up with a mod of a unit like that to get rid of the noise, it would be blindingly obvious and would show clearly on test results. Other mods probably produce very subtle changes, and given that in real world conditions, mic selection and placement etc will have a much bigger effect on the end result, one has to suspect that we're talking about owner satisfaction rather than practical outcome. And of course owner satisfaction is something most of us here pour quite disproportional amounts of our hard earned cash into!