Strange simulation behaviour

Hi,

as a part of my project concerning the usage of ns2, I've frequently
stumbled upon simulations that performed poorly without any straightforward
reason so I decided to ask someone to maybe explain to me what was going on
and how can I correct the problem. My original TCL script creates a random
topology and communication scheme, but I extracted an instance where weird
things happen. Here are the files:

demo.tcl
<http://network-simulator-ns-2.7690.n7.nabble.com/file/n31102/demo.tcl>
demo.scn
<http://network-simulator-ns-2.7690.n7.nabble.com/file/n31102/demo.scn>
out.awk
<http://network-simulator-ns-2.7690.n7.nabble.com/file/n31102/out.awk>

The TCL file has two variables at the very start: cbr_scenario and
pck_scenario. When both equal 0, the simulation generates a PDR of 18.26%,
but when either or both equal 1, it jumps to about 99%. The only thing the
variables control are the CBR packet size (70 vs 100), and communication
start times.

This seems rather arbitrary. There seems to be no concrete reason for the
result of one simulation to differ so much from the other three, and when
generating random simulation scenarios this posses a problem. I find this
kind of behavior far too frequently.

So, can anyone  explain what is going on?

Thanks,
Leo