For some reason, the link you posted doesn’t seem to work. I assume you are referring to this one. If so, you are correct, the graph represent the absorption per unit volume for each cell.
In a 2D simulation, we only assume the structure and fields are invariant in the z direction, so the result should be interpreted in the same way as in 3D. We could do a 3D simulation using periodic boundary conditions in z and we would get the same results for each z plane.
By default, the fields recorded by the monitors are normalized to the source spectrum. In linear systems, it means the response doesn’t depend on the source spectrum (same if the source had a constant spectrum over all wavelengths) as discussed here.
In some application (such as solar cells), we are interested in the response of the system to a specific spectrum (solar spectrum for instance). Instead of running the simulation with a custom source that would reproduce this spectrum, we use the linearity of the system and simply apply the spectrum to the normalized results. The advantage is you can use the same simulation with different spectra, without having to rerun the simulation.
By dividing Pabs by the source power, we obtain the fraction of power absorbed. It means we can easily calculate the power absorbed for any source power illuminating the device. Again, we can modify the illuminating power without having to rerun the simulation.
I hope this helps! Let us know if you have any further questions.