Blair,
I don't think that there are general issues with imbalance, but the solution of course gets more difficult in the phreatic case due to potentially large parameter contrasts. I'd recommend to turn on the error norm visualization (in the context menu of the User Data section in the Data panel) to see whether there are regions in the model with a large remaining error. Having idenified those, you might have a better insight into potential solutions.
Good luck!
Peter