As far as I know, ULA's track record for reliability is much better. Last I read, maybe a few months ago, SpaceX launches had failed 1 in ~19 times, and ULA's failed zero in ~85. I wonder how the Air Force addressed this issue. Perhaps the cost of failure is less than the premium for reliability; perhaps they evaluated reliability based on something other than the track records (e.g., an engineering review); and of c…
It's very easy to address even if you assume 5.5% failure rate will remain constant and not decline Space X is still significantly cheaper.
Google Terra Bella (aka Skybox) uses SmallSat (~100 Kg) imaging satellites, and the NRO could use a similar for priority uses where a launch failure wouldn't be as painful as loosing a $1Bn spy satellite. A SmallSat wouldn't be as capable as the larger ones, but has advantages.
The purpose of the EELV program is more about assured access to launch vehicles to get to space, than individual missions being near 100% success.