That sounds like a bug on our side, but it also just sounds like a bad idea to try to do this without using a document that is set to the CE prediction function you used in generating your original optimization methods. That is kind of an assumption of the code, at least until you get everything stored in an optimization library. It is not really valid to tell Skyline you want to use "None" as your CE prediction function, and Skyline should refuse to accept that outright.
You need to use the same CE prediction settings you used to generate your original methods for the 50 injections. All that gets stored in the resulting data files is the offsets from the original predicted values. If you do tell Skyline what you used to predict those values, then it will do the math for you and give you valid CE values in your exported method.
Again, if you want to forever store the actual CE values without needing to refer to the original predictor, you should build and optimization library. For now, you need to be careful about the steps you follow to make sure you don't go setting the CE predictor to something else or definitely not "None" between when you exported your optimization methods and when you try to export your optimized method.
Hope that helps to clarify. Thanks for explaining your situation.