Elsevier allows you to post the preprint, but not the final published version (which includes Elsevier's typesetting, logo, etc.)
In case of (a), I actually prefer to post my own PDFs, because one of the tasks publishers-as-middlemen generally perform is screwing up captions, misapplying 'style guides' and so on.
Regarding (b), they are clearly trying to make it more difficult to find these posted articles, and to reduce the number posted by ensuring that individual authors have to take the initiative (rather than it being a matter of institutional policies).
(Edit: reading your comment again, I see that I'm wholly agreeing with you).
"For subscription articles: These rights are determined by a copyright transfer, where authors retain scholarly rights to post and use their articles."
http://www.elsevier.com/journal-authors/author-rights-and-re...
Absolutely not. Copyright is spit into several separable rights and the authors have signed some of these away. That they still own the majority of their rights under copyright is irrelevant.
> and have always had the right to distribute their papers on their personal websites
No they haven't. This was allowed (for preprints) under Elsevier's T&Cs only a few years ago. Authors have signed away their right to publish their work, so they need explicit permission.
You may not like the agreement, but it's perfectly valid under the law.