The closest analogy to FSL that I would recommend in AFNI would be to eschew the polynomial regressors in AFNI and instead put in trigonometric regressors to pick up the low frequencies you want filtered out, and let the linear regression + pre-whitening model do it all at once." So by eschewing polynomial regressors you mean to change the polort option to 0, correct? But I do not understand how to put in trigonometric regressors to pick up the low frequencies. Is that what 3dTProject accomplishes?
3dTproject removes the unwanted components from the data, and that's it. It is intended for use with resting state data, where there is no a priori "signal" component.
For task state data, you can generate the trigonometrical time series that correspond to a range of frequencies using the 1dBport program. Then these are input to 3dDeconvolve (and hence 3dREMLfit) via the -ortvec option. The examples in the output of 1dBport -help should provide some guidance --
http://afni.nimh.nih.gov/pub/dist/doc/program_help/1dBport.html