Hi,
Unfortunately I do not have a good answer for you, I just wanted to say I had a similar situation, and was trying to find out the details of how the computations differ between the two versions.
In my case, the disparity was resolved when a coworker reviewed my model and noticed that erroneously high mannings roughness had accidentally been applied to the bridge area; when that was resolved the disparity in model results was eliminated. However, we still do not know exactly why this was the case (as the version 6.1 results also used the ‘incorrect’ geometry file and produced different results). Based on my observation, it seems that the updated version is more sensitive to roughness, so you could try looking at that. Sorry I can’t provide a more complete answer.