Why implement iradon with a filter which is designed directly in the frequency domain?

In the iradon documentation it is written that the filter of the backprojection algorithm is designed directly in the frequency domain and then multiplied by the FFT of the projections. It is done so despite the fact that in the book of Kak: Principles of computerized tomographic imaging, (which is referenced from the iradon command documentation), it is stated that a better approach (that eliminates reconstruction artifacts) is to multiply the FFT of the projections by the FFT of the impulse response of the filter. I wanted to ask why you do not use the better implementing, which Kak recommends ? (I ask it since I have non negligible errors in iradon reconstruction of even a circle image)