Three basic ingredients are required to generate a simple stellar population (SSP) library, i.e., an initial mass function (IMF), a stellar evolution model/isochrones, and an empirical/theoretical stellar spectral library. However, there are still some uncertainties to the determination and understanding of these ingredients. We perform the spectral fitting to test the relative parameter offsets between these uncertainties using two different stellar population models, two different empirical stellar libraries, two different isochrones, and the Salpeter and Chabrier IMFs. Based on these setups, we select five SSP libraries generated with the Galaxev/STELIB and Vazdekis/MILES models, and apply them to the pPXF full-spectrum fitting of both MaNGA and mock spectra. We find that: 1) Compared to the Galaxev/STELIB model, spectral fitting qualities with the Vazdekis/MILES model have significant improvements for those metal-rich (especially over-solar) spectra, which cause better reduced $��^2$ distributions and more precisely fitted absorption lines. This might due to the lack of metal rich stars in the empirical STELIB library, or code improvement of the Vazdekis model. 2) When applying the Vazdekis/MILES model for spectral fitting, the IMF variation will lead to not only a systematic offset in $M_*/L_r$, but also offsets in age and metallicity, and these offsets increase with increasing stellar population ages. However, the IMF-variation caused metallicity offsets disappear in the case of Galaxev/STELIB based libraries. 3) The Padova2000 model provides a better match to the MaNGA galaxy spectra at [M/H]$_L-1.0$. Current tests suggest that spectral fitting with the Vazdekis/MILES+BaSTI combination would be a better choice for local galaxies.

19 pages, 17 figures, accepted for publication in MNRAS