Could the original files be at different sampling rates?
Only a guess but perhaps they did some at 11k or 22k and others at 44k depending on what recording quality they needed, then set each playback speed in whatever file defines the sounds (resource.bin, config.bin?? not sure).
Which, assuming that BIS did this, would mean that you must set yours at the same rate to match. or a custom definition of your sound files.