In the body of the text of Tragedy of Great Power Politics, M. says that "the only assumption dealing with a specific motive that is common to all states says that their principal objective is to survive...." (p.32) But because "there are many possible causes of aggression" [what are they? he doesn't say] "and no state can be sure that another state is not motivated by one of them" (p.31), assuming that all states all the time want nothing more than survival is not warranted. Indeed, in an important end-note -- why this material is buried in an end-note rather than being in the text is rather perplexing -- M. makes clear that:
Security concerns alone cannot cause great powers to act aggressively. The possibility that at least one state might be motivated by non-security calculations is a necessary condition for offensive realism, as well as for any other structural theory of international politics that predicts security competition. (p.414 n.8)Again, he doesn't say what these "non-security calculations" or motives are [except for a brief discussion on pp.46ff.], but this is nonetheless an important clarification. It's one that tends to get lost later in the text, however, for example at the beginning of ch.6 when he writes that "security considerations appear to have been the main driving force behind the aggressive policies of Germany, Japan, and the Soviet Union" in the twentieth century (p.170). This, of course, is in some tension with the statement in the end-note that "security concerns alone cannot cause great powers to act aggressively." It may not be a fatal logical contradiction; it's probably more the result of careless use of shorthand phrases (and, to be fair to M., I have quoted only part of the end-note here, not the whole thing). Still, someone who only reads the text of TGPP and doesn't read the notes is likely to be even more puzzled about this issue than someone who has read the notes.
Harknett and Yalcin, in their 2012 article "The Struggle for Autonomy," which I discussed in this post (where, I see, I also quoted the Mearsheimer end-note I've quoted here) recognized some of these problems about state motivation in realist theory and attempted to deal with the issue more satisfactorily than had been done previously. Although I was critical of their effort (not that I have gone back and carefully read that post), they should be given credit for having recognized and tried to address the problem.
Added later: For one useful discussion of this set of issues (and, of course, more thorough than the discussion in this post), see Jack Donnelly, Realism and International Relations (2000), ch.2, "Human Nature and State Motivation."