There's an article about voting theory in the
March Scientific American.

Two parties makes sense when the choice criteria
are clustered along one axis. And, a third party
should be able to break off from the middle in an
orthogonal direction when enough voters are
undecided between the two majority parties. The
third party candidate is not a spoiler when they
siphon equivalent numbers of votes from the established parties.

There is no purpose in complaining about the Greens
(or any other party on an extreme end of the major axis)
not catching hold. They are so far from a majority that their
appropriate role is simply to tug at the candidate closest to them.

The strongest third party in the US that is not
built around a cult of personality is the Libertarians.
But their message is too alien for most voters
(although they all seem to have personality.)

I guess the real question is, does the two-party
system artificially project diverse interests
onto a single left-right axis? Would better
policy result instead from dynamic coalitions for every issue?

The advantage of clustering by strong parties is that the process
of developing the party position entails negotiation and compromise that forces
some coherence among interests prior to facing the other party.