Compared with experimental spectra, calculations for conjugated phenyl and thiophene oligomers tend to overestimate the ground state Raman intensities of higher-frequency vibrations (1200-1800 cm(-1)) relative to the intensities at lower frequencies (<1200 cm(-1)). The discrepancy was observed in previous benchmarking work that examined the method dependence of the calculated Raman spectra for a series of aromatic molecules. This paper further investigates the nature of the discrepancy by examining the role of anharmonic corrections and the dependence of the calculated Raman spectra on the inter-ring torsion angle for the representative molecules biphenyl (BP), 2-phenylthiophene (PT), and 2,2'-bithiophene (BT). Perturbative anharmonic corrections to the spectra calculated using density functional theory (DFT) provide only slightly better agreement with experiment. On the other hand, calculations at larger torsion angles give up to 30% improvement in the relative Raman intensities compared with the spectra calculated at the optimized geometries. The torsion-angle dependence of the Raman intensities is most pronounced for delocalized C-C and C-S stretching modes, and less pronounced for bending and ring distortion modes that do not involve inter-ring stretching. Higher-level calculations using the coupled cluster with single, double, and perturbative triple excitations [CCSD(T)] method indicate that DFT underestimates the energy barrier for torsion isomerization at small angles, and it overestimates the barriers at large angles, thus predicting minimum geometries at torsion angles that are too small. Therefore, the results suggest that the discrepancy in relative Raman intensities may be related to an overestimation of inter-ring conjugation by DFT, which also tends to favor geometries that are too planar.