The accuracy of the time-dependent self-consistent-field (TDSCF) approach assuming partial factorization of the total wave packet is tested against an exact treatment, when applied to calculate asymptotic properties. The test is carried out in the framework of a three-dimensional simulation of the Ar-HCl UV photodissociation dynamics. All the partially-separable TDSCF ansatzs possible for this problem are investigated. The quality of the TDSCF results is found to be strongly dependent on the specific partially-separable ansatzs applied. In general, the TDSCF predictions are in very good (even quantitative) agreement with the exact ones for magnitudes associated with direct photodissociation dynamics, and are qualitative in the case of indirect photodissociation. The deviation of the TDSCF results from the exact dynamics is interpreted in terms of an error operator defined as the difference between the exact and the TDSCF Hamiltonians. The analysis of this operator also explains the different accuracy of the partially-separable ansatzs investigated. Based on this analysis, a simple procedure is suggested to estimate the relative average quality of the different TDSCF ansatzs.