Isn't that what the I (integral) term is supposed to compensate for?
Here is the way I see it. A Pure PD (proportional, Differential) controller makes smaller corrections as you approach the set point. When you get inside of a certain band the corrections are essentially zero, so you will have an offset. The integral term is designed to compensate for this offset by summing (integrating) the error over many control cycles. To set the integral term you need to let the process run to determine the offset then set an appropriate integral term. It seems like everyone who does PID controllers use slightly different terminology, it would be a lot easier to set up PID if they could standardize the terminology.
The offset is when a system under PID (Proportional Integral Differential) control stabilizes a small distance from the control set point. So if you were using a temperature PID controller with a set point of 100 deg, the system would stabilize at 98.