Dumping what you have and going for a rebuild mid development is almost always a bad idea.
For a personal project, I did exactly this. It was originally built during the betas of asp.net MVC. I got the app to a stage where it was usable (actually I still use it daily), but it was nowhere near ready for the outside world. And this was the problem; it was going to take an enormous amount of work so that other people could use it...
When Silverlight 3 was announced, I literally grabbed the backend of the app - stuck RIA services in between and had a few screens up and running that day without any prior SL knowledge. I probably could have kept going down this path but something clicked when I started to realise the power of silverlight. The goal posts for my app moved, and I began a SL specific rewrite.
Since then, I've started re-writing about 5 times over. I guess I'm still just learning how to best build an app in SL, having spent the last 12 years or so of my career working on stateless web apps, there was a big mental shift involved.
I'm a much better web developer then I am a silverlight developer, but if it was for a real project (rather then a pet side project) - it would have been shipped and out the door by now.
I'm convinced that SL is the ideal platform for most web applications (as long as it being a plugin isn't going to be any issue).
With that said, shipping is still the most important thing. SL is great, but the learning curve is steep. If you guys are anywhere near completing the app, I'd insist you forge on with mvc and maybe get someone to build a SL branch.