Thermodynamically, the capsule has the effect of a thermal low-pass that is in front of the low-pass that represents the thermal inertia of the logger. This causes a 2nd order response (instead of a first order exponential function).
In simple terms, there is a short delay between a thermal step function applied and the result recorded.
In addition, the total time constant (staying with a simplified approach, just ignoring that it is a 2nd order differential equation) goes up considerably. The DS9107 data sheets specifies for the thermal response time constant in Note 3: "The value was derived from submerging an iButton capsule with DS1922T inside into 55°C water. The initial temperature of iButton and capsule was 23.5°C. Without the capsule, the response time constant is approximately 30s." With the DS9107, the thermal response is five times as slow as it would be otherwise. The capsule does not electrically connect the iButton inside to the outside world. A capsule made of metal would allow for a much shorter thermal response time, because metal conducts heat much better than PPS.