"The challenges are that you'd need to develop modules for orbital assembly. As the ISS experience has shown, this isn't easy. If you launch the modules on unmanned vehicles (like the Russians did with the Mir modules), then each module would need it's own TT&C, power, propulsion and guidance systems to be able to fly itself to the assembly point. All of these systems make each module more expensive and complicated and, once the vehicle is assembled, represents dead mass."

It depends on the mission and the architecture. You've forgotten that use of a common system would reduce R&D.

Speaking of dead mass, for about 60% of the US, manned Mars flight represents dead mass.

My thinking for 7 years has been that the only thing that may save manned space at all might be the Comet and Asteroid Protection System. It's certainly not manned Mars, or a race with China and/or Russia simply to get to the Moon. As that race is probably already lost, why run in it?