To be quite honest I find the biggest drawback for HTML5 on mobile devices is speed. Even simple applications have a noticeable lag that isn't present (or can be hidden) on Native applications. As an example (as used in the article), Facebook is one of the worst performing apps on my phone, often taking me to places I never touched and crashing occasionally. Therefore I would suggest that the hardware/implementation still needs to catch up first before choosing a blanket HTML5 architecture (which mind you the hardware is doing gradually).
Now, perhaps it because I am testing on older hardware; however this is for good reason. I test based on hardware that my clients/target audience use - if they upgrade and they get a speed increase then that's a bonus - but until then they need to get the best possible experience given the hardware they have. I will use HTML5 for applications if I can get a reasonable return on investment by doing so, but usability is always number one - if it doesn't perform well then it may as well be thrown out.