Firstly I think either approach could work, but you maybe should think about things like how your are going to maintain and communicate the state of the system across days.
For example, if I return on a set day to complete the final piece, do I get taken straight to my final step?, or do I review the previous data?
I think if the previous steps are set in stone, making these two tasks separate would make more sense. That way it's a bit like the process of applying for something, then returning to check on progress / approval, when I can submit finally.
Other questions to ask might be, can the user have more than one instance of the set of steps active at one time? If so, the previous sentence probably still holds as it might afford the design the opportunity to display a list of their incomplete work, and the dates they need to return.
Alternative, if the step process is a single thing the user does, and they don't do anything else in the system, perhaps taking them back to their last page with clear indicators as to the state of their wizard could work well.
One thing I'd do either way is make sure when they reach the end of what they can do on a given day, that it's clear they are finished. Nothing worse than not knowing if you are meant to finish the wizard or not, making a sequence which can't be ended would be frustrating.