I know it's a chicken and egg problem, you want a large enough sample base to get users using the product but with vague messaging you are attracting people that wont use your product (people that you shouldn't target) and people that WILL use your product but don't understand the vague messaging. I would suggest taking all versions of the page and signup methods to test usability of the product. Optimizing on the initial signup is great, but if your users don't use the product after they signup bc it's so vague.. then you've just wasted all that optimization effort. Unless the goal is to create a huge list of signups.
Also, where is traffic coming from? Paid, organic search, social media, press release, tc, etc? That has a huge difference on your results. And what tools are you using to run the tests and what is the sample size? I think readers would love to hear those tidbits too
Thanks for sharing your findings.. good luck =)