This is what the current software renders for the word "tira": <img width="598" alt="current_incorrect_rendering_of_tira" src="https://user-images.githubusercontent.com/4258357/29252696-de807f4c-8039-11e7-9505-7f94f996bbe5.png"> But "i" and "a" should belong to the preceding consonants "t" and "r" respectively. Should render something more like this: <img width="619" alt="proposed_rendering_of_tira" src="https://user-images.githubusercontent.com/4258357/29252702-ed802920-8039-11e7-9001-1198e9e7cd95.png">