The JVM performs more minor GCs (which was something to be expected, since the new generation now has less heap space), but the total minor GC time has not increased. Also, it performs much fewer major GCs, and while the time for each full GC is a little longer than before (which was also expected, since the heap space allocated to the old generation has increased), the overall time spent in major GCs is less.
For a more striking example of this phenomenon, see the example in my "101" paper. Actually, in that example I avoid full GCs altogether!
You really should also total the amount of time spent in GC over the course of the run and determine the percentage of time spent in GC. That is one of the data points I have in my latest white paper that I am presenting at CMG in two weeks. I am also using R to analyze the data, beats Excel!