The shear strength parameters of loess samples are determined from conventional triaxial shear test results and used in the rational design of sustainable geotechnical infrastructures. However, the rubber membrane that is used in the triaxial shear apparatus for applying the all-around pressure to the test specimen has a significant influence on the measured shear strength parameters. In this paper, remolded and undisturbed unsaturated loess samples from northwest China are used in a comprehensive testing program to determine the shear strength from triaxial tests and understand the influence of a rubber membrane. The results show that the measured undrained cohesion from unconsolidated undrained triaxial tests on unsaturated soil specimens with and without a rubber membrane are significantly different. In this study, differences in the shear strength with and without a rubber membrane are assessed from shear strength index values that can be determined from undrained cohesion and the internal friction angle derived from conventional triaxial tests. Experimental results suggest that predominant changes arise mainly in the undrained cohesion values. The change rate of shear strength indices values of undisturbed loess shows a strong correlation with its water content; however, it is weak for remolded loess. The correlation coefficient between error and measured values of all shear strength indices is more than 0.8. Empirical correction relationships for triaxial shear tests with a rubber membrane for three different types of loess were established from the investigations. The simple approach used in this study can be used as a reference to apply corrections to the measured undrained cohesion values of unsaturated loess samples from northwest China.