The reliability of the classification varies due to a number of constraints. Basically, the object classes are defined by features of the objective prism spectra and the direct images, and are therefore not coincident with common definitions. Also the object classes do not need to be complete in themselves. For example, although, the class "QSO/AGN'' contains most of the quasars and Seyfert1 galaxies, part of them will be found among "galaxies'', and many of the fainter optical counterparts are still unidentified. To make a more quantitative completeness test we have correlated the HRC with the most recent Véron-Cetty and Véron AGN catalogue (Véron-Cetty & Véron 1996) and found 539 coincidences. 481 of them were recognized on our plates as AGN candidates, 2 were empty fields, 16 classified as galaxies, 3 as clusters of galaxies and for 37 AGN no classification was possible based on the HQS Schmidt plates.
Among stars, the assignments of spectral types are fuzzy with respect to the usual Morgan-Keenan classification. Early M-dwarfs may have a classification as M-dwarf (Class 5) or as K-star. Similarly there is a gradual transition between the F and G star subclass and the K-star subclass.
In principle, the catalogue is also affected by confusion. There could be objects inside the error circle which can be X-ray emitters from their optical classification, but actually are not the real counterparts in these specific cases. Except for bright stars the surface densities of the potential X-ray emitting object classes are so low, that this problem is negligible. This confusion was also discussed in Bade et al. (1995) and we refer the reader to this paper for more details. A recalculation of the numbers given in Bade et al. (1995) for the 3847 HRC positions with an error radius of 20'' let us expect less than 1 serendipitous QSO in the HRC, but up to 8 "bright stars'' with B < 12. Since the average error radius is smaller, these values are upper limits. Therefore some of the "bright star'' classifications (Code 7) can be erroneous. Note that for many of them the plausibility of the identification cannot be checked, because saturation inhibits an estimate of the spectral type. Another possible source of contamination is proper motion of nearby galactic X-ray emitters. For our identification project we used nearly exclusively plates taken within five years of the X-ray RASS observations. Even with proper motions of several arcseconds per year this problem is negligible.