I know that this is Stata Forum (not SPSS)! But I came across a very interesting discussion that may be interesting for Stata users, as well, see: paired t-test for umatched subjects impossible to compute, correct? (it is not easy to follow because it is a mailing list discussion). Btw, it also shows that Stata is easier to use than SPSS when you want to simulate data.
It is not that I want to (re)start a discussion here that is already ongoing elsewhere, but the issue itself is interesting enough that you may forgive me pointing to a discussion in the SPSSX-list in this forum -- as for me, it will be an exception from a general rule.