No matter what, the Modi is async and will be reclocking the signal.
There is no doubt in my mind that the data is bit-perfect, not a mere attempt… otherwise I wouldn't be able to flawlessly play DoP from iTunes.
I am not familiar too much with the architecture of CoreAudio on the Mac. Most of what I said I am passing on from an engineer who is more familiar with this aspect of OS X. However, I do have a good idea about device drivers and how system services are presented to the software running on the computer, in this case an audio player. So I am making an educated guess here. I would not be surprised that integer mode does bypass some of the processing involved, because this approach should not require additional conversions between different types of number formats before the end result is passed to the DAC. But I doubt a system setting like this one will bypass the entire audio processing stack of the Mac OS. That is why I use the term "layer", meaning only some software of the OS is probably bypassed. And this is why I like what is termed as "direct mode" and "integer mode" on Aurdirvana. Supposedly Audirvana goes directly to the DAC and feeds it numbers in integer format.
Would this result in audible differences? According to one EE that I know tells me that the bypassing of CoreAudio completely will have audible benefits. But he has not addressed the benefit of integer mode with me. Perhaps one more familiar with CoreAudio should comment here about the intricacies of this part of Mac OS. Who knows. I may have put the proverbial foot in the proverbial mouth. :-)
By my minimal understanding having seen CA from a dev perspective, that seems an accurate explanation, I would only say that the 'processing involved' should just be additional, complicated math. I'll have to look into what's gone on w/ CA in 10.9, I still run 10.6.8, and CA has always been a rock-solid stack for pro a/v needs (which are, of course, rather different than audiophiles' needs at times…!)