The current pythia setup, as far as I can tell, is the following.

Pythia 6.2
Perugia tune
No manual setting of <kT>, which probably means <kT>=1 GeV
MSEL=1 

Two jets are considered "matched" based on the distance in eta-phi space, with cutoff of 0.5. 

With this setup, we get ~50% matching between detector jets and particle jets. This seems to be largely driven by the fact that Pythia filter is uniform in phi, whereas the detector is not.

Fig. 1. Kinematics for all particle jets found in the pythia sample passing the filter, and the detector-particle jet matching efficiency

The matching between parton and particle jets are also low, and it is strongly pT dependent, suggesting that we have a large contribution from underlying events. (previous study)

Fig. 2. RED: Unmatched particle jets, BLACK: Matched particle jets
 

Even at high pT, we only get ~50% matching, which seems too low to me. 

What I propose, for now, is the following.

Pythia setup

Version: 6.4
Tune: TuneA
Gaussian width of primordial kT = 0.5 GeV  MSTP(91,1)  PARP(91,0.5)

ckin3=0.0
ckin4=-1.0 

 
MSEL=0  //manual selection of processes