I think another key point is that often the government tags a much higher cost on "risk of failure" when evaluating proposals.

Startups are fast, but they aren't always reliable. At least for values of reliable that == being able to throw money at {Large_Technology_Contractor} to have them parachute in an additional team of engineers to solve unexpected problem X. Or being around 10 years from now for support.

But agreed, there are all levels of involvement. Consider the prime contractor's cut off the top the price of getting bid into the game and contact with the right people, then do your best to work your way up!