I'm not familiar with Chromsword, but did look into Drylab, Fusion, and ACD Autochrom a couple years ago (so some of my recollections might be out of date). They all have their good points so it really depends on what kind of information you want from the software. In other words, how do you expect the software to help you with Quality by Design?
Drylab and (I believe) ACD use parametric modeling (they fit the data to a theoretical model) whereas Fusion uses empirical modeling. Parametric modeling should give you better models in fewer runs, but it's a big question in my mind how many real world separations don't fit the theoretical models for one reason or another. I think empirical modeling with Fusion is more flexible and can give you a more rigorous statistical analysis but it takes more work and the model may not be as good as the best parametric model. The modeling done by Drylab seemed very “black box” to me, in contrast to Fusion which is very transparent. From what I remember Drylab could not interface with Empower so all sample sets and instrument methods had to be built manually (but I think it can interface with some other CDS'). Fusion can do all of that automatically (with some limitations) and that was a big selling point. I think ACD had project management and knowledge management tools that looked good but I never actually tried them out.
I doubt any approach works 100% of the time. S-Matrix claims their Rapid Development scheme works 80% of the time. I don't have enough experience to confirm or deny that, but I would be surprised if Drylab or ACD worked significantly better than that. The questions to ask are how much those 80%'s overlap and which one includes more of your typical separations. I believe they all offer free trials.
Hope that helps.