The answer is simple, they are two different sources or error and you need to add both together to understand the total accuracy or distance to a known point.

The device gives you one, a horizontal accuracy number based on its calculations to where it thinks it is in the 3D reference frame. The OP’s question related to this calculation to better understand the accuracy.

However, the device also transforms that 3D position into the 2D horizontal coordinates of where it thinks it is in your chosen local datum. But it’s often actually somewhere else because the transformation is simplistic, and it does not calculate additional plate movement over time. The known point is actually further away.

So that additional error needs to be added to the device provided accuracy number to give you the total distance to the known point if you want to understand the real accuracy of your position.

So if the question was how does it calculate the accuracy, then one short answer is that it does it insufficiently because it leaves some of the error out.