If I understand what the OP has done... it "shouldn't" work!
Don't confuse shouldn't work with not optimal. Per the Guidelines for Reliable Long Line 1-Wire® Networks at Dallas http://www.maxim-ic.com/app-notes/index.mvp/id/148
the star topology is listed as a 1-Wire network topology...
Star topology: The 1-Wire bus is split at or near the master end and extends in multiple branches of varying lengths. There are slave devices along, or at the ends of, the branches. ... When different topologies are intermixed, it becomes much more difficult to determine the effective limitations for the network. As a rule, the designer should apply the most conservative of the criteria in these cases.
Note, the document does go on to caveat star topologies...
Precautions with Star Topologies
Testing has shown that unswitched star-type network topologies (i.e., those with several branches diverging at the master) are the most difficult to make reliable. The junction of various branches presents highly mismatched impedances; reflections from the end of one branch can travel distances equal to nearly the weight of the network (rather than the radius) and cause data errors. For this reason, the unswitched star topology is not recommended, and no guarantees can be made about its performance.
For my application I was less concerned with "guaranteed performance" from Dallas and more concerned with simplicity in building and deploying. My network currently consists of about 200 feet of cable split between 5 nodes on a star topology and has been working fine for months.
If you are looking at deploying a network with many devices over 100 meters or more then you probably want to stick to a linear or stubbed topology...