Accuracy and uncertainty analyses are essential for every measurement technology. In crystal orientation indexation by electron backscatter diffraction (EBSD), a series of accuracy estimations have been made for the Hough transform and dictionary indexation methods. The mean angular deviation is a standard parameter to indicate orientation accuracy, but this criterion is indirect and closely related to the accuracy of the projection center coordinates. Precise known orientation relationships are necessary to evaluate orientation accuracy without the ground truth. The current work uses the natural crystal twins and hardware orientation relationships to assess the orientation accuracy directly. The accuracy level for different EBSD analysis methods is compared through four experimental data sets of varying pattern definitions and noise levels. It is found that the full pattern match (FPM) algorithms improve the accuracy as compared to Hough indexation, and the gain varies greatly between 14% for fast acquisitions and 20 times for high-quality patterns. Depending on the resolution and quality of diffraction patterns, FPM results in an accuracy of crystal orientation between 0.04° and 0.9°. Comparing the two FPM variants, matching the gradients of diffraction patterns performs better in the case of high-to-median quality acquisitions while matching the pattern itself is more accurate for more noisy and low-definition patterns.