Servocam,
I understand that we all respond to applications here looking through a keyhole, not being enough information to adequetely analyse the problem, but I have to disagree with your statement.
For example, say the reflected inertia matched the motor inertia in the original system, so servo wise it was perfect. So by going higher on the motor the system was out of match 2:1. Then net system, if it had the same torque as the original, would respond slower.
Taking a motor of equal torque and half the original motor inertia, the system would respond faster.
And with todays modern controls, it should handle a larger mismatch in either direction and still perform. A 5:1 mismatch doesn't seem to bother any of the servo people I have worked with lately.
Given this data, I would still go with the same torque, lower inertia, motor option, and feel more confident it could handle the application.