Helle Wayne,
I think the slope in Fig. 37 fits well to a second order filter: take e.g. the orange curve (+/-5V): at 40kHz I read -15dB, at 80kHz I read -25dB, which is better than 6dB per octave.
I think problem with insufficient attenuation at Nyquist-frequency is more related to the position of the corner frequency: with 15kHz (or 23kHz for +/- 10V) it's just 2-3 octaves away from 100kHz, and a second order filter has no chance to reach -90dB. If you need this attenuation, I think you really need an additional external filter.
best regards
Achim