It is true that the input match will be disrupted by the changing impedance. Question is if you actually care ...
What really matters is the power that is transmitted into the air. To know this you can model the dipole as a series RLC circuit. The R value is the radiation resistance, the L and C are defined by the Q factor of the dipole. These can be approximated by the measurements you made.
Now if you drive this parallel circuit with a port with source impedance R, the maximum power is dissipated into the antenna and transmitted in air at the resonance freq. Outside the resonance frequency, the antenna is less efficient.
Now with the lowpass filter in the middle, the antenna will still radiate less efficient. You should now evaluate the S21 with the RLC circuit for the antenna instead of the simple R to find out the suppression due to filtering and antenna efficiency. Then you should evaluate the spurs that your transmitter generates and whether these don't violate any EMC rules.