Details of the algorithm are in the JGR paper available here. But a brief description of some of the limitations that users should be aware of are included below.
If aerosol levels are too high the retrieval will not be able to match the measurement vector and extinction is not reported. This has a tendency to low-bias the measurements immediately following volcanic eruptions, particularly at lower altitudes. There is no hard limit, but generally saturation occurs near extinction values of 0.01 km-1, although this depends on both altitude and scattering angle with low-altitude backscattering conditions saturating soonest.
Although clouds have been flagged as best as possible, distinction between aerosol and clouds is never-the-less an ongoing issue in the UTLS region.
While biases due to particle size have been reduced in version 7.1, the retrieval is still somewhat sensitive to the assumed lognormal distribution, and this can alias into the retrieval as a function of the solar scattering angle (SSA). As the SSA is a function of latitude and time, this can lead to biases that correlate with both parameters.