Yes, I could deal with the rest of them, but #3 really reduces its usefulness. It's still better to find a publicly accessible paper on academia.edu than nowhere, but having to page through it online and not being able to download a PDF is not good, especially when the authors in most cases didn't intend that restriction.
The arXiv is my favorite of the preprint repositories, but focuses mainly on scientific areas. For the social sciences, law, humanities, etc., SSRN (http://www.ssrn.com/) is a decent alternative. Some people are wary of it because it's for-profit (like academia.edu), and it also has a bunch of pointless cruft (such as a completely useless categorization system and proliferation of "top-10" lists), but it does at least allow you to download the PDFs. It's really improved access to legal scholarship especially; like with physics/arXiv, law/SSRN is now a pairing that's widely known in the field and it's basically expected you'll upload your preprint there if you want to be read.
I don't get the link, so yes, apparently one has to register. Personally I find that quite horrible. There are serious efforts underway to open up research, and putting stuff into yet another walled-garden (or rather, dozens of different walled-gardens) is a step backward.
Consider that the people that put their stuff their would ordinarily perhaps have put it on their personal homepage or some e-print service, so comparing it to the $20 fee of publishers is not the right comparison.
The download option is shown to logged out users too, so you should see it. But you do have to be registered in order to compete the download.
With regard to your other point, we think that the conversion of the PDF to HTML 5 and a couple of other things we do on Academia.edu massively increases the visibility of your paper as compared to hosting it on a department webpage.
This is a huge benefit for the individual researcher in increasing the impact of their work and the number of people who find and use it.
The arXiv is my favorite of the preprint repositories, but focuses mainly on scientific areas. For the social sciences, law, humanities, etc., SSRN (http://www.ssrn.com/) is a decent alternative. Some people are wary of it because it's for-profit (like academia.edu), and it also has a bunch of pointless cruft (such as a completely useless categorization system and proliferation of "top-10" lists), but it does at least allow you to download the PDFs. It's really improved access to legal scholarship especially; like with physics/arXiv, law/SSRN is now a pairing that's widely known in the field and it's basically expected you'll upload your preprint there if you want to be read.