refinement of a structure i done with X-ray diffraction data and with coordinate geometry constraints. The structure is constrained to theoretical bond distances and angles and also because the number of x-ray terrns, for 1 Angs data, is much larger than number of atoms the refine coordinate errors can be less that 0.01 Angs. Imagine taking and average of 10 readings, the standard deviation of the average is proportional to 1/sqrt(10) =0.31 and for 100 reading goes down to 0.1. So the more X-ray and geometry term used the smaller the coordinate error becomes. So for X-ray data the refined accuracy of the model can be much better than data resolution. The reason why the electron density does not have this high resolution when it is viewed is because you are seeing the data resolution and not resolution after refinement,

Hello Miroslav. Thanks a lot for your answer.