Gaussian: How can I hold initial guess symmetry?

Dear all,
     From the Gaussian Manual (Gaussian 94 User's Reference P.88-89), we are
 able to model an excited state by transposing orbital with Guess=Alter.  And
 it recommends use SCF=QC or SCF=DM in order to hold initial guess symmetry
 > from one iteration to the next.  But I found that even using these two
 keywords, the initial symmetry is not able to be held for some molecules.
 So my question is:
 Are there any methods such that we are able to hold the initial guess
 symmetry with 100% certainty?
     I am a new to the computational chemistry.  Any kinds of suggestions are
 useful.  Thank you very much in advance.
 Kiniu WONG (Kin-Yiu)
 Undergraduate student
 Department of Physics
 The Chinese University of Hong Kong
 Email: kiniu %! at !%  or  kyiwong %! at !%