The issue of the correlation of AC loss data obtained in different experimental configurations for high-temperature superconductor (HTSC) samples of different topology is investigated using a practically important example of a hollow cylinder with induced current. The basic configuration under consideration is a transformer device where the current in a BSCCO cylinder is induced by the magnetic flux from a coil positioned inside the cylinder. Using a novel contactless method, the AC losses in the cylinder are measured. It is shown that Bean's critical state model is unsuitable for explanation of the experimental data. Moreover, in the case of complete penetration, the AC loss values appear to be quite different for a slab, a hollow cylinder in uniform external magnetic field and a cylinder in the investigated configuration. Thus, neither the critical state model nor the experimental data obtained on samples of simple geometry can be used to determine AC losses in HTSC hollow cylinders. A method for the AC loss evaluation in HTSC cylinders that is based on using the real E-J characteristic and the mathematical model of a transformer is proposed. It provides a good agreement of experimental and calculated values of AC losses.