The first approximation made is to assume that there is only one light path.
In reality the photons collected by the instrument have followed very different light paths through the atmosphere, depending on the interactions they have had with air molecules, with the surface of the Earth or with clouds.
However, for most situations we can assume that it is enough to consider the processes along a single 'effective' light path. The length of this light path is dealt with separately using airmass factors derived from radiative transfer calculations.