I'm a bit surprised at the opposition to modding this deck, when many people here have other modded Marantz and Edirol decks from Oade and Busman.
I would defend to the death your right to mod the device if only with the aim of knowing - or believing - that you're getting the best possible result from it, whether that can be proven or not, and, whether it actually matters or not. After the mod, you may be totally convinced it's an improvement from subjective listening, but you'll never be able to hold that up as proof of an improvement. That's not to say that it won't be interesting to hear your opinion in due course.
Perhaps out of ignorance, I find it hard to believe that recording amplified loud music reveals very much about subtle differences between recording devices or mods thereto, apart perhaps from their ability cleanly to record the high SPLs involved! The sound would be more heavily influenced by the quality of the amplification of the instruments. (See also later comments re mics).
When it comes to classical music (my thing), as I've said before, I don't believe that there's a lot to be gained in reducing system noise level below a certain point. Quiet classical music should be reproduced quietly (otherwise, at the ear, the frequency response will be skewed and then it isn't a high fidelity reproduction of the original). So you shouldn't be turning up the replay level to the point where system noise is heard (unless you've seriously messed up the recording level at the outset or your recording device is really naff).
Nature recording is a different thing - sometimes you might want to reproduce ambiences at a natural level, so as with classical music, you shouldn't be turning up replay to the point where system noise is heard. But sometimes your intent is to make audible that which is too quiet or too far off clearly to be heard, and then system noise can be more of a factor.
As for frequency response, I would have thought that any half decent preamp in a recording device these days will have an essentially flat frequency response across the audible range. If a mod to the device adds (for instance) some kind of presence boost in the higher frequencies, while that might sound more sexy, it's no longer accurate. And you can fiddle with the response with eq in post production anyway. I rather suspect that if I posted some samples purporting to be from a 70D before and after a modification, and actually just added a slight HF emphasis to one set of samples actually coming from an unmodded device, most people would identify that as coming from the modded device, simply because it would "sound better".
At the end of the day, what makes the biggest difference in the recording chain is the mics and their placement. Before spending money on a mod it's worth considering whether it would be better to spend on better mics. Of course, if the cost of the mod is less than would buy an improvement elsewhere in the chain, go for it. Then save up for more costly, but probably more effective, improvements in the rest of the chain.