I suggest Waze adds a property to each tts term, allowing editors to choose, for example, between 'Houston [pronunciation 1 {hues-ton}]' and 'Houston [pronunciation 2{house-ton}]' and so on. One region could default to [P1], while another could default to [P2] but exceptions could be specified.
I'll bet that a lot of what is needed to support the idea is already part of Waze because of the different languages supported. This could be an expansion of that support.