To calculate ALOGP and AMR, the algorithm assigns each atom in a molecule to a particular atom type defined by Viswanadhan et al., 1989. However, the code appears to have a bug in the assignment of atom type so not all atoms in a molecule will be assigned to an atom type. This causes errors in ALOGP and low values for AMR.
I have attached two files containing the same molecule but with different aromatic representation. For the first file (000095-14-7.mol), the algorithm is unable to assign the middle nitrogen in C-N=N to an atom type. For the second file (000095-14-7_aromatic.mol), it is only able to assign two atoms (one hydrogen and one nitrogen) to an atom type. Thus the AMR is vastly under-estimated for the second file.