Correct. I ended up going another route. Rather than creating a publication for each object type, I did the first method instead where I created one publication for all objects and changed the parameter for -SkipError to skip errors where it couldn't find the embedded object. Once synchronization is complete, I rerun synchronization without skipping errors. Now since the object it needed is now in the Subscriber database, it replicated any object it skipped in the previous run. Not the cleanest method, but it made everything easier with one publication. Since I am replicating 9 databases, I did not want to create multiple publications per database. That would have been a nightmare.