When establishing the feasibility of matching experimental and theoretical convergent-beam electron diffraction (CBED) patterns to recover structure-factor information it is important to investigate the accuracy of the results obtained and the effects of random and systematic errors. The effects of counting noise in a digital data-collection system are studied, together with various sources of systematic error. The sensitivity of parts of the pattern to specific structure-factor parameters is analysed via the calculation of analytic gradients of the intensity with respect to those parameters. All calculations are based on diffraction at the  axis of GaP, using simulated CBED patterns as idealised "experimental" data. The results indicate that, provided care is taken to minimise the effects of systematic errors, low-order structure factors may be extracted from experimental data with an accuracy approaching 0.1%.