The "illusion" you refer to is called doppler shift. It can be corrected if you know the correct inputs, but we have to guess because we don't know the velocity of the car, nor the exact angle-of-travel relative to the recording source. If you're making a relative comparison, and the inputs are close enough, you needn't correct for it.
I used the 1:30 region for analysis.
Samples:
- Pre-shift: 01:28.2, 00:00.3 duration
- Post-shift: 01:29.2, 00:00.3 duration
Highest amplitude peaks (not corrected for doppler, Hanning window):
- Frequency pre-shift: 289 Hz
- Frequency post-shift: 232 Hz
Plots: