hm I just thought maybe I shouldn't do this because I only did the imputation because Amos was complaining that it can't calculate certain things when there are missings in the file. But this way I'd have the missings again. So I'll just calculate the things for AMOS only with the imputed file without the control group. 

I'm not very clear on why variables are missing in each group, and why you are including a covariate in your model that you didn't ask one of the groups. My gut reaction is that you shouldn't do something to one group that you don't do to the other, and this would include imputation. I also feel that if something is missing for the whole group, it sounds dangerous to impute for it. I think missing by design is quite a different thing from partially incomplete data by whatever mechanism. These are just some random thoughts, and feel free to ignore if they are unhelpful - I may have completely misunderstood the purpose of your study. Good luck.

Dear Annika, I do agree with what Darren already metioned. However you may still want to report on some characteristics that only valid for either of the groups. 

If you want to limit the imputation to a certain subsample, you can use "if" and explain the criteria for impuation. 

Regards, Mehdi