Jaehoon, you want to experiment with output port power and Miller modes. I don't remember if you can change Miller modes in SessionOne/PowerSession, but you can through FX9500 web GUI. There is no single best recipe, since it will depend on tag density and environment, but the number of combinations it finite :) Overall:

The greater the power - the greater the range, but you may run into reflection/coexistence issues, which will negatively affect Rx signal quality.
The lower Miller mode you use - the greater the range (but speed falls down, just like with Wi-Fi Rates/MCS).

Typically, the best combination for any particular setup is found by experimenting. P.S. Ensure corrent antenna orientation and take cable length (attenuation) into account as well. :) P.P.S. These and many other topics are covered in RFID Design & Deploy training (AEL2300).

I would like to offer the following clarification: In any given environment, the read range of an FX9500 should be purely a function of the transmit power.  This is based on the assumption (and fact) that the range is "forward-link-limited".  This should be the case certainly using Miller-4, and even using Miller-8 (which theoretcially has more reverse link margin) or Miller-2 (which has less link margin that Miller-4 or Miller-8).  It could be possible that if you use FM0 (no symbol redundancy) then the link might become "reverse-link-limited", and therefore, you could observe reduced read range using FM0 as compared to Miller-4. I hope this does not confuse the matter more as it gets into some technical depth that honestly no-one should ever need to worry about.