[Minitab] DOE Model Output vs. Actual Data

#1
I am running multiple 3 factor DOEs and each one of my model outputs fits very well, all above an 80% fit. The problem I am having is that when I plug our recorded values into my model for parameters x,y and z my model outputs a significantly different/unfeasible value when compared the the actual recorded measure.

Although, when I check using response optimizer I get a value that does resemble the actual recorded output.

I know this is somewhat vague, but my question is regarding if there are any simple fixes or anything that I might be missing in my DOE that would cause my model to appear to have a great fit but not be able to predict values close to my real values.

Thanks