You need to do a cost:benefit analysis, with the following questions in mind.
How many users are on the various platforms? (Related: Are you losing sales because you don't support other platforms?) What are your projections for how this will change?
What are the development cost for each your three scenarios?
If you move from a desktop application to a web-based application (either traditional or desktop-class on the web), what functionality (if any) will you lose? If you lose functionality, how will your users handle that loss?
Developing for the web doesn't mean that you must support all browsers and all screen sizes. You can still specify that you will only support specific browsers (say, Internet Explorer and Firefox) with a minimum screen size/resolution. Doing this means that you don't automatically get a mobile web application, but that could be added in a future version as you get more comfortable with web development.
In my employer's case, we went from a very rich enterprise desktop client developed in C# and moved it to a desktop-class application on the web. We've mostly had to use Flex because HTML5 doesn't provide everything that we need so that we didn't lose important features. We're closely tracking HTML5 development, but at this time, Flex is the best option for us. We had a web version of the old desktop client too, but it only had a small fraction of the features of the desktop client, and was rarely used. Moving to a desktop-class application is allowing us to grow our support for OSes (we chose not to support every OS and every web browser on the first release of the desktop-class web application). This might or might not be the best solution for you, but it was for us.