Edit: Adamsmith beat me to it, my post basically replicates his, the last section should still be relevant though, unless I have the wrong end of the stick.

For starters, the only way I can even find to add a locally stored article to my library was by clicking the Add Link to Local File button in Zotero's main window.

PDFs in Zotero are generally stored within the Zotero data directory itself (unlike in say, Endnote). To use Zotero's full capabilities you will need to import your existing library of PDFs into the Zotero data directory. To do this you can use the "store copy of file" menu item under the "new item" button (plus sign in a green circle). This shoud allow you to select all the PDFs you want in zotero (ctrl clicking to select multiple files) and import them all at once.

For the next bit you would need to be running Zotero 2.0 beta (in my opinion it is pretty much as stable as 1.0 if you don't use the synchronisation features, and quite stable enough for everyday work even if you do). In 2.0 beta select all the PDFs you have just imported and select "retrieve metadata for PDFs". This searches Google scholar for information on the PDFs and automatically associates them with full bibliographic metadata. Note that you would probably want to do this at a rate of not more than a few hundred PDFs each day - otherwise you will fall foul of a security policy in Google scholar which prevents excess automated requests for information and you will be blocked from Google scholar for a few hours.

If you already have your PDF library organised in a folder structure your best bet is to replicate this using collections (left hand pane) and import into these collections a folder at a time (files will import into the currently selected collection).

Once you are happy that everything is safely in Zotero, and that Zotero is the right tool for you you should be able to safely delete the PDFs in their original location if you want to (note that keeping a regular backup of your Zotero data directory is always a good idea)

My impression is that it was able to scan this document and parse a bibliography since academic articles almost always use a consistent format for storing references.

This feature "create bibliography from selected item" doesn't do this (although I agree the error message does somewhat imply that it does). Instead it allows you to select a number of items and make a bibliography of those items which you can then paste into a document or email etc. (when actually writing an article/paper/book though you would be better advised to use the word processor plugins, "create bibliography" is a rather quick and dirty method). You are presumably using this tool on your links you created to external PDFs, which will fail as there is no metadata associated with these standalone links with which to build a bibliography.

I don't believe that you can parse the bibliography from a PDF itself with Zotero, and I am having difficulty seeing why you would want to do this, at least in the biological sciences (my field) this would just result in a huge list of articles, most of which would have only marginal relevance to the topic of interest. To store specific references of interest from an article you are reading you can find these online (using google scholar, web of knowledge, scopus, the publisher's website etc. etc.) and grab them using the "page" or "folder" shaped button whcih should appear in the Firefox address bar when you are looking at the article page or a search results page. Alternatively you can copy and paste the DOI, ISBN or PMID number into the "add item by identifier" box (magic wand button) and Zotero will attempt to find the appropriate article/book.

a) How is it generating the metadata? I don't have any PDF metadata stored in the file but it's able to figure it out somehow.

See http://www.zotero.org/support/retrieve_pdf_metadata for more information. Basically it looks for a DOI first, and uses that to find the metadata online, if it doesn't find one it queries Google Scholar using the text from the PDF (i'm not sure exactly which text makes it into the query though).

b) Sometimes it figures out the article metadata, but it is completely and totally wrong and has nothing to do with the actual article. Is this just a known bug that I have to deal with?

This is to do with the fact that Zotero just grabs the top match from the Google scholar search. There has been talk of allowing the user a choice of several matches, but I don't know whether this is firmly on the cards (sorry, can't find the relevant thread ATM). There is a thread here where you can report these issues to help the developers improve metadata retrieval.

c) Sometimes it just fails on certain files and says "No matching references found." How can I manually do what it's doing and create this top level document node with a link to the PDF under it?

This happens when there are no matches in the Google scholar search. You have two choices in this situation:

1) Find the paper/article online and click on the icon which appears in the address bar to automatically save metadata, then drag the PDF onto this item (if a copy of the PDF isn't automatically saved - for this to happen you may need to activate the appropriate option in the preferences).

2) click the "+" icon, choose the appropriate reference type, and drag the PDF to the new blank item, then fill in the metadata fields by hand.

There has been talk of option 2 happening automatically if metadata retrieval fails.

as for the last question: I assume you mean for an entire issue of a journal?
That would work if the journal is online with one of the common publishers or JSTOR, EBSCO etc. - you would then go to the issue of the journal and click the folder in the address bar of Zotero and select all - that should do.

As for the first question: You can't reload, no - I take it you have found the magic wand where you can add an item by identifier (including the DOI).

no -doesn't look that way - at least not easily.
Apparently there is a way to "save an inssue to a binder" when you have an ACM account -
If you could export that to Endnote (and have the automatically import RIS checked in the Zotero General Preferences) that _might_ do it for you. But let me emphasize _might_ as I haven't tried this out.