I have done a lot of work with EFDC and think it's a wonderful tool. I have also worked with John Hamrick, the developer of EFDC. I have used both the rectangular and curvilinear capabilities and ran the first validation of this for John using the laboratory data of Ye & McCorquodal obtained with an experimental test facility consisting of a 270° bend more than 20 years ago. I have attached a side-by-side comparison of the rectangular and curvilinear model results (zoom in to see the vectors). I increased the resolution of the rectangular grid until it produced reasonable results. Which option you use depends on what you're trying to model. This link shows one of many EFDC models that I have http://dudleybenton.altervista.org/projects/Diffusers/index.html
Tecplot is a great tool to use in conjunction with EFDC. This is an EFDC model of a branched river system. http://dudleybenton.altervista.org/projects/Diffusers/export.gif