On point #2: describe your product in clear, what-it-does language.
Mistakes I see are emphasizing: how it does it (C, Java, OO, Rails, REST, ...), where it does it (PC, mobile, Mac, Cloud, ...), "ecosystems" it integrates with (Social, FB, Oracle, ...), who your investors or team are (VC, founders, investors...) etc. All of which may or may not be particularly relevant, but ... they're not key to me understanding what you do. Tell me these things, but focus on the what first.
Use direct, actionable language, not vague or nebulous terms. It's a "NFS file security permissions auditor", not "Cloud information assets security tool".
Describe a workflow or workflows from the perspective of your users. Not developers. Not architects. Not
This doesn't just apply to startups. I use a lot of Free Software, and many of these projects also fail to describe themselves clearly (though most, especially over time, eventually get it right, if only because other people can come in and rewrite idiotic descriptions). Reading through a list of package descriptions from Debian or Ubuntu, where a pithy, one-line description is your shingle to the world, should give a sense of good and bad descriptions.
Even long-established technologies such as Java suffer from this.
At www.java.com we have "What is Java?": "Java allows you to play online games, chat with people around the world, calculate your mortgage interest, and view images in 3D, just to name a few. It's also integral to the intranet applications and other e-business solutions that are the foundation of corporate computing." Um. OK. open http://www.java.com/en/download/whatis_java.jsp
At Oracle, we have a Java landing page with ... no description of the technology or its components (which aren't self-evident): http://www.oracle.com/technetwork/java/index.html
One of the best succinct summaries I've seen in recent memory is from jwz's "Java Sucks" page:
there are four completely different things that go by the name "Java": 1. A [programming] language, 2. An enormous class library, 3. A virtual machine, 4. A security model. http://www.jwz.org/doc/java.html
Now that is something I can wrap my head around (he also goes on to describe strengths and weaknesses of each component, good essay, read it, it's still disappointingly relevant).
Note though: the best product description comes from a critic. If you fail to clearly define yourself, your critics will.