Spin Density vs SOMO surfaces

 Dear all,
 I wanted to visualize the localization of the unpaired electron
 in radical-cations whose structures were optimized at the UB3LYP/6-31G*
 level in Gaussian98.
For that purpose, I asked Gaussview to compute both Spin Density and SOMO surfaces. 
 In most cases, both pictures were very similar. However, in some cases,
 I was surprised to observe two very different pictures, i.e.
the surfaces were located over completely different regions of the radical-cation.
In some cases, alphaSomo, alphaSomo-1 and betaSomo were found very close in energy. I tried to obtain a "mean view" of the system by adding the alphaSomo and alphaSomo-1 
 then subtracting betaSomo (with the cubman utility within g98 package).
 The resulting picture remained different of the Spin Density surface.
 Has somebody an explanation?
 All the Best
 **********************************************
 Dr Pascal Gosselin
 Laboratoire de Synthhse Organique/UMR-CNRS-6011
 Faculti des Sciences & Techniques
 Universiti du Maine
 Avenue Olivier Messiaen
 F-72085 Le Mans Cedex 9 France
 Tel: + 33 (0)2 43 83 33 71
 Fax: + 33 (0)2 43 83 39 02
 Email: Pascal.Gosselin.-at-.univ-lemans.fr
 **********************************************
Soutenez le mouvement SAUVONS LA RECHERCHE : http://recherche-en-danger.apinc.org/ 
 **********************************************