A summary of haemoperfusion targeted more at the time-poor CICM exam trainee is available in the Required Reading section for renal SAQs in the Part 2 exam.
Random resin or activated charcoal particles
There is a certain selectivity; the molecules need to penetrate the pores of the porous particles in order to get adsorbed; and so you can adjust the selectivity of the sorbent by altering the size of the pores. This wont get you far. All molecules will still be adsorbed as usual, but only a certain size of molecule will be able to access the full surface area, so proportionally, that molecule will be selected for. And any molecules smaller than that. So, its not particularly selective.
Ligands, immobilized in a matrix of sorbent
Apart from straightforward unselective adsorption (which still occurs) there is also some antibodies (or whatever other ligands) embedded into the filter, which clean the blood of their specific protein. For instance, ant-TNF-alpha antibodies can be embedded onto the surface of little microbeads
For a definitive treatment of all of this, you ought to pay homage to the gigantic and all-encompassing "Critical Care Nephrology" by Ronco Bellomo and Kellum (2009).
The Gambro and Fresenius websites have also been an excellent source of information.