When group members are directly responsible to each other, as in the case of a small startup or egalitarian cooperative, this effect disappears.
I'd imagine it would be fairly straightforward to model this in game-theoretic terms.
This is something of a hot topic in the economics/psychology are now - several new papers are out along the same lines, but most of them are still paywalled. I have a hunch that predispositions to act one way or the other will turn out to be strongly correlated with a person's perceptions of social mobility.
A large-scale (albeit stereotypical) example of what the OP is talking about is in the economics of labor relations, where an individual 'eager beaver' attitude and the collective goals of a union are somewhat at odds.
Not sure about that egalitarian cooperative counterexample. As shown whenever people are co-opted into a cooperative, "all animals are selfish, but some animals are more selfish than others."
However, your startup example is certainly in line with other research suggesting cooperation among kin:
http://svenssonresearchlaboratory.blogspot.com/2009/04/under...
Perhaps the key is the social networking mentioned in this link. These students in this submission's article were not necessarily part of a social network. This has implications for the HR department when a company grows large enough to have a "company culture".