I have a dataset with 678 taxa. There are no heterozygotes so I replaced every double character (AA) with a single representative character (A). Total there are 40843 characters. I need to conduct a parsimony analysis, but a simple heuristic search with NNI and default parameters takes a long time and gives really bad results (obviously). What is the best way to perform a parsimony analysis on a large dataset without taking days that will give me decent results?