In my experience, using Speakable Items for the rather, shall we say, complicated applications can get a bit tricky - but there's no reason it's not possible. Usually, if you use the "Make This Speakable" command whilst an application is in focus (i.e is your foreground app), a folder will be created for it in Application Speakable Items if it does not already exist.
As OS X is UNIX-based, my initial idea was to call the folder eclipse, as the executable that the system sees (not the .app bundle itself) is actually called eclipse as opposed to Eclipse.
However, should that not work, Apple's official advice is to simply drag the folder to your desktop and then attempt to use the "Make This Speakable" command whilst using (in this case) Eclipse so that the computer will (or at least should) create the folder for you - at which point you can simply drag all the AppleScripts you have made into the newly-created folder.
From a programming / behind-the-scenes point of view, though, the answer may well hinge on what version of Java you are using! Should you be using Apple's own Java implementation, you will likely run into more problems than using the latest Java 7 release.
Eclipse uses the Standard Widget Toolkit (SWT), which wraps native controls (or more accurately, makes Java communicate with OS X like it would with, say, other *nix) so by those standards alone ought to already be accessible. I know that NetBeans is mostly, if not entirely accessible at least from a VoiceOver standpoint.
Browsing Eclipse's documentation and bug history reveals discussion of this back in 2007, and seems to suggest they were working on it in 2008, but there's no mention of having Eclipse controlled with AppleScript voice commands. It ought to be straightforward and simple, so if it seems like Eclipse and/or OS X is resisting every attempt to communicate with Java or the IDE itself via AppleScript(s), I'd be inclined to believe it's some quirk with Eclipse or its communication with the SpeechWorker and the rest of OS X.
According to that link, it affects only speed, duration & offline ability, not accuracy.
It downloads the database locally to your machine, allowing the Mac itself to do the work, rather than having to send it to Apple.
Presumably, though it's not mentioned, the learning accuracy will still improve over time, as it learns from its mistakes.
Best Answer
Check out the Voice In Speech to Text Chrome Extension. It builds on Google's speech to text to allow you to dictate directly on many websites like Gmail, Wordpress, Zendesk, Facebook, etc. Note: it works only in the Chrome browser.
You can download from the Chrome Webstore