Characterizing natural fracture systems involves understanding fracture types (faults, joints, and veins), patterns (orientations, sets, and spacing within sets), size distributions (penetration across layering, aperture, and trace length), and timing relationships. Traditionally, observation-based relationships to lithology, mechanical stratigraphy, bed thickness, structural position, failure mode, and stress history have been proposed for predicting fracture spacing along with the relative abundance of opening-mode fracture versus faults in fractured rocks. Developing a conceptual fracture model from these relationships can be a useful process to help predict deformation in a fractured reservoir or other fractured rock systems. A major pitfall when developing these models is using assumptions based on general relationships that are often site specific rather than universal. In this paper, we examine a mixed carbonate-shale sequence that is cut by a seismic-scale normal fault where fracture attributes do not follow commonly reported fracture relationships. Specifically, we find (1) no clear relationship between frequency (or spacing) of opening-mode fractures (joints and veins) and proximity to the main fault trace and (2) no detectable relationship between fracture spacing and bed thickness. However, we did find that (1) the frequency of small-displacement faults is strongly and positively correlated with proximity to the main fault trace, (2) fracture networks change pattern and failure mode (extension versus shear fracture) from pavement to pavement through the mechanically layered stratigraphic section, and (3) faults are more abundant than opening-mode fractures in many areas within the fracture network. We interpret that the major fracturing initiated near maximum burial under relatively high-differential stress conditions where shear failure dominated and that mode-1 extension fracturing occurred later under lower differential stress conditions, filling in between earlier formed shear fractures. We conclude that whenever possible, site-specific observations need to be carefully analyzed prior to developing fracture models and perhaps a different set of fracture network rules apply in rocks where shear failure dominates and mechanical stratigraphy influences deformation.