Bayesian methods have become a standard tool in the retrieval of exoplanetary atmospheric spectra. An “ignorance prior” is normally used for the chemistry of the atmosphere to preclude any constraint before observation. We assert that the common paradigm of parameterizing the chemistry as a spectrally active trace species and an inactive filler gas (such as H2/He or N2), successfully applied to atmospheres of hot Jupiters, should no longer be relied upon for atmospheres with high mean molecular weight, as it places a biased prior on the abundance of the background gas. As a solution, Benneke & Seager (2012) suggested using symmetric, simplicial priors, such that no gas is assumed to be background and all included species are treated equally. We test the validity of the common paradigm by comparing full retrievals with the trace & background parameterization and with symmetric priors. We find that using the former approach can result in unjustified or even spurious constraint on the retrieved abundances of the background gas in certain situations. We also find that the Rayleighan slope of the spectrum in the optical and UV range plays an important role in constraining the abundances of spectrally inactive gas. We explore degeneracies with other effects, including aerosols and disequilibrium chemistry.