The force backwards on the motor mount during the coast phase is an internal normal force, incapable of pushing it out. The normal force is simply what is keeping the motor from moving forward through the airframe. Draw a free-body diagram of the rocket as a whole and of the part.
I've flown a rear ejection rocket before, my Giga Drill Breaker. It was a loose fit: if I held the airframe (the cone), the motor mount would fall out easily. Every single flight, it stayed together until the motors' ejection charges went off. There could not have been anything pushing the motor mount out during the coast phase, or else it would have come out.
If a rocket were in a vacuum and is coasting starting at 98 m/s, each second it will slow down by 9.8 m/s due to gravity. However, since rockets are not launched in a vacuum, the rocket airframe tries to slow down more than that due to aerodynamic drag. There is no aerodynamic drag on the completely-internal motor mount, so it tries to decelerate at "only" 9.8 m/s^2 while the airframe is slowing harder. This difference in external forces on the two parts results in a normal force between them but it is only able to keep their accelerations the same: if their accelerations are the same, they do not separate.