QSAR - Help for Ghose ALogP

Dear Colleagues,

we are involved in a deep check of the AlogP calculation performed by DRAGON software.
However, it seems that the AlogP values published in the Ghose paper of 1989 (JCICS, 29, 163-172, 1989) are not reproducible for some molecules.
On the other side, the atom type definitions in the paper are not completely univocal.
We attach some molecules (MDL mol format) and we hope that some of you can give us some more explanations and/or the atom type partitions of the molecule atoms.

The attached molecules are (with the ALogP Ghose values):

no. 360 - 4,5-dibromo-1,2,3-triazole (ALogP : 1.07)
no. 1211 - cytosine (ALogP: -0.55)
no. 1367 - 2-methyl-2-imidazoline (ALogP: -0.55)
no. 1754 - 2,4,6-trichloropurine (ALogP: 2.89) - HyperChem format
no. 3899 - sorbose-6-phosphate (ALogP: -0.97)
no. 7539 - 1,3,6-trimethyl-5-(dimethylamino)-uracil (ALogP: -1.14) - HyperChem format
no. 8993 - antipyrene (ALogp: 1.54)
no. 9820 - vitavax (ALogp: 0.09)

Thank you for your help

 Roberto and its group

--------------------------------------------------------------------------

      Prof. Roberto Todeschini
     
      Milano Chemometrics and QSAR Research Group
      Dept. of Environmental Sciences
      University of Milano-Bicocca
      P.za della Scienza 1 - 20126 Milano (Italy)
      http://www.disat.unimib.it/chm/