The pixel-to-pixel spectral fitting technique is often used in studies of stellar populations. It enables the user to infer several parameters from integrated light spectra such as ages and chemical abundances. In this paper, we examine the question of how the inferred parameters change with the choice of wavelength range used. We have employed two different libraries of integrated light spectra of globular clusters (GCs) from the literature and fitted them to stellar population models using the code STARLIGHT. We performed tests using different regions of the spectra to infer reddening, ages, [Fe/H], and [{alpha}/Fe]. Comparing our results to age values obtained from isochrone fitting and chemical abundances from high-resolution spectroscopy, we find that: (1) the inferred parameters change with the wavelength range used; (2) the method in general retrieves good reddening estimates, specially when a wider wavelength range is fitted; (3) the ideal spectral regions for determination of age, [Fe/H], and [{alpha}/Fe] are 4170-5540, 5280-7020, and 4828-5364{AA}, respectively; (4) the retrieved age values for old metal-poor objects can be several Gyr younger than those resulting from isochrone fitting. We conclude that, depending on the parameter of interest and the accuracy requirements, fitting the largest possible wavelength range may not necessarily be the best strategy.