Hi Alex,
we are working for eliminating the various all.js instances as you suggested...
Now the strange stuff:
for preventing users to access content without login...we use...Joomla access system...!!
and to let user register and access we use...JCLogin from JFB Component...!!

I took an article:
/NACIONAL/investir-em-educacao-e-saude-de-criancas-e-atacar-desigualdade-na-raiz-destaca-dilma.html
and changed access level from "registered" to "public" (from article panel)...
and worked! the og:image is correctly displayed.
Your component give a full user access / registration that perfectly integrates with Joomla user management system, using facebook login...
Is it possible that it does not work with content that are just for registered users?
Am I the only one that have this big issue related to images?
The issue is blocking because if the crawler/script (that one which is responsible for scan an article, find the first image, and create the og:image meta property) does not work when a Joomla article has access level set to "registered", it makes no sense to use that automatic function.
In fact if I want an user to login /register, it's because there is something in terms of content that is just for him (as registered).
This is a function that was born with Joomla (set an article permission as registered).
Moreover, the fact that the crawler does not work over content for registered user, create a cascade of negative effects:
1)when a user post a comment over an article for "registered" users, on his facebook wall is published a totally random image;
2)on reccomendations module, images associated to articles are wrong;
3)on every aspect of the Component that involves images, that images are wrong....
4)even if a user login, the image is not correctly shown in comments/reccomendations/etc...
I think we can start investigating from the fact that other parameters, such as "title" are correctly fetched from article (also registered), the only thing that is not is the image (og:image). I think it's something with your code, but I don't know, really...
Now: the milion dollar question: how to solve? Put a manual reference to the og:image for each article will triplicate the workout.
If you explain to me which script/files are involved into the process of scanning pics from articles, I can ask our engineers to take a look on it, maybe we can also help to find a solution.
Thanks, Antonio