I would say that the main points are performance, access to hardware apis, and - the most ignored one - ease of install.
There is no familiar way for users to install a web app (unless they are a bit tech savvy). My uncle adds favorites to its home screen to go to web pages, he knows that, and its impossible for him to think that a website is a website and if you add it to the home screen "turns out" to be like an application.
If a clearer interface was provided in the mobile phones, something like identifying webapps with a meta tag, or something, so that the phone could know and offer the user to install the app, the barrier for mobile webapps adoption could be lowered a lot.
I agree that "installing" an offline web app on something like the iPhone is a bit odd - largely because there is very little you have to do. You can just install it as a Bookmark or on the Home Screen (where it can have it's own logo etc. - like a normal app).
I'm more targeting getting structured (or semi structured) data onto a mobile device, possibly make the content editable and allow changes to be sync'ed back to the server to be shared with others or collected via an export or API.
e.g. I can take an Excel spreadsheet, turn it into an offline HTML 5 app, allow users to edit the data, sync all the changes and export it back out as a modified spreadsheet.
There is no familiar way for users to install a web app (unless they are a bit tech savvy). My uncle adds favorites to its home screen to go to web pages, he knows that, and its impossible for him to think that a website is a website and if you add it to the home screen "turns out" to be like an application.
If a clearer interface was provided in the mobile phones, something like identifying webapps with a meta tag, or something, so that the phone could know and offer the user to install the app, the barrier for mobile webapps adoption could be lowered a lot.