The objective of this study was to examine the effect of pulling angle on time-zero mechanical properties of intact infraspinatus tendon or infraspinatus tendon repaired with the modified Mason-Allen technique in a canine model in vitro. Thirty-six canine shoulder samples were used. Twenty intact samples were randomly allocated into functional pull (135°) and anatomic pull (70°) groups (n = 10 per group). The remaining sixteen infraspinatus tendons were transected from the insertion and repaired using the modified Mason-Allen technique before being randomly allocated into functional pull or anatomic pull groups (n = 8 per group). Load to failure testing was performed on all specimens. The ultimate failure load and ultimate stress of the functional pulled intact tendons were significantly lower compared with anatomic pulled tendons (1310.2 ± 167.6 N vs. 1687.4 ± 228.2 N, p = 0.0005: 55.6 ± 8.4 MPa vs. 67.1 ± 13.3 MPa, p = 0.0334). For the tendons repaired with the modified Mason-Allen technique, no significant differences were observed in ultimate failure load, ultimate stress or stiffness between functional pull and anatomic pull groups. The variance of pulling angle had a significant influence on the biomechanical properties of the rotator cuff tendon in a canine shoulder model in vitro. Load to failure of the intact infraspinatus tendon was lower at the functional pulling position compared to the anatomic pulling position. This result indicates that uneven load distribution across tendon fibers under functional pull may predispose the tendon to tear. However, this mechanical character is not presented after rotator cuff repair using the modified Mason-Allen technique.