Divergent results for different mesh in convection

Hi All.

I want to calculate heat flux parameter.
So.. conditions of task is next.. I have floor with some temperature and ceiling with some lower temperature.. conditions on walls is adiabatic.
At first I decided to figure out mesh parameters to optimize accuracy goal and time of calculation. But after some sequence of simulations I realized, that results aren't convergent. They didn't bound to a certain value with higher mesh. Please see picture below and shared source files on google drive:

Google drive Links
HexagonalDome.SLDPRT - Google Drive
Simulation data with different mesh parameters: https://drive.google.com/folderview?id=0B3l02HJw4lTycjZuNzl1UkxZSzg&usp=sharing

Thanks for your attention.
Regards, Oleg.