I disconnected my 4.7K pullup resister and the DS18B20 reading changed from 25 to 127. Restoring the connection to the resister restored the reading to 25. The sensor board is wired for parasite power so I cannot easily test if the pullup is needed if powered directly from +5V.
I picked up an RJ12 cable from Radio Shack this afternoon so I could bring the sensor to the bench. It came with the plugs preattached. The lower temperature readings work fine without the pullup resister so I guess it depends on how you plan on using the sensor. For my case where I want to detect temperatures that approach freezing I do not need the pullup resister. To get readings much over 50 I had to add the pullup that the Maxim spec showed in its diagram.