Well this reinforces the fact that in-camera downsizing is crap (hence the softness) however I disagree on both points (detail and noise).
First of all, downsizing inevitably reduces noise by the very nature of the process. You are getting rid of pixels, and that means noise pixels too. So it's not exactly fair to say "look - downsized files are better" because it's not an apple-to-apple comparison. Yes you have less noise, but you're giving up pixel information (detail). Downsizing also has the same effect on any sharpening you do - if you downsize, you loose sharpening. And if your downsizing algorithm is really crap, you even lose some of the original (out-of-camera) sharpness too (which is evident from your samples). This is why sharpening is always done as the very last step in the workflow. But I digress...
I think you are mistaking perceived contrast with detail. You keep referring to how "clear the separation between books is" but look at the text on the books. There is way more detail there simply because you have more pixels to display them. The contrast on the black line is misleading.