In recent decades, several computational models have emerged at the mesoscale for predicting the failure development and strength of composite materials. Nevertheless, the intricate failure mechanisms of composites pose a formidable challenge in developing models capable of consistently reproducing diverse sets of experimental data. Furthermore, the absence of standardised multiaxial tests has hindered consensus on failure envelopes and criteria for composites. In this work, an in-house constitutive model implemented in a finite element code is employed to predict the failure strength of several filled-hole compressive carbon/epoxy experimental tests, varying in stacking sequences, sizes, bolt preloads, and other parameters. The influence the selected failure criterion has on the predictions is analysed. The findings indicate a large impact of the failure envelope on both the failure compressive strength and developed failure mechanisms. Therefore, the correct adjustment of the failure criteria can be as relevant as defining accurately the material input parameters. These results emphasise the importance of accurately establishing failure envelopes, rather than overemphasising the development of complex models.