If total BW is defined as the 0.707 amplitude threshold ( -3dB) and on the lower half BW averages to -35kHz at 965 KHz, we might approximate it as a BW of 70kHz with a Q= 1000/70= 14 or inversely 1/Q=7.1%
Given a 10% 90% rise time,T = 0.35/f(-3dB) we expect a burst envelope rise time of 0.35/70kHz = 5 us or approximately 5 cycles at center f.
The center f tolerance is 30kHz @ 975kHz or 3%. The actual error is 25kHz or +2.5%
The effective rise time will not change significantly when a pulsed carrier is within the 3% tolerance spec. The load impedance and steady-state power load will be very frequency sensitive off actual center, yet not during the risetime of the pulse.
BW is not the same as Tolerance and is always expected to be greater otherwise it must be manually tuned.
Other ( for the advanced user)
e.g. a Crystal with a natural 100 ppm tolerance is expect to have a BW of at least 100 to 200 ppm while MFG's with better process controls can achieve 25 to 50 ppm tolerance at 25'C although the BW will likely be the same.
Again %Tolerance must always be less than %BW.
However in a "good" part, the %error tolerance on the %BW relative to centre f or Tolerance error of Q will always be much less than the center frequency error .
Special 10MHz SC cut Xtals with <0.00001 ppm (1e-11) error at some ovenized temp, will still have a phase noise BW greater than 1e-11. These are used for extremely stable OCXO's and Stratum Level clocks.
A very high Q LC filter with a % BW of 0.5 or a Q of 200 but has a tolerance error of 1% is not very practical as the Q is greater than its centre tolerance with poor results. Since this is more difficult to produce low tolerance in LC components than crystals, LC BW's < 1% are difficult unless manually tuned and then must be temperature compensated.
How will YOU define your requirements?