IMHO, the 2-angle configuration was simply derived through mechanical tinkering. It seemed to work and so was subsequently measured and documented -- but history tends to live on.. Since the DVD-grating spectra tend to use only a portion of the camera's image "pixel width", there is some "flexibility" in the configuration which will still show a spectrum.
The downside is the asymmetry at the sensor which adds non-linearity to the spectral wavelength - to - pixel relationship so the 2-angle configuration is less than ideal. IMHO, the grating should be parallel to the camera sensor plane and generally within a couple mm of the lens. Such an assembly is then mounted at 45-deg to the incident light. For some graphical explanations see: https://publiclab.org/notes/stoft/03-02-2015/plab-3-spectrometer-upgrade-prototype .