Hi
It depends on what stage you are validating the method to support.
For Phase 1, there are no set-in-stone recovery guidelines, most people use accuracy and linearity . I would just bracket your dose and perform three replicates of each to show recovery. Of course, if you use validation as time zero of a stability study, you would run all three.
For Phase 3, (which I don't do very often), you do need to show recovery (including spiked recovery) at least at all three levels. You may also have to extend that range to 80-120% of the known doses.