When MSA from T-COFFEE was used all mutations were predicted to be neutral,from CLUSTA-W only few mutations were predicted (2/56)but msa from COBALT predicted results for all the mutations.
The alignGVGD program builds its analysis directly from the MSA. The three programs you listed all use different algorithms to create a MSA and therefore will almost surely produce different alignments. Clustal-W is an outdated program and does not produce nearly as good of alignments as its replacement Clustal-Omega. I would go to pubmed and find an article comparing the accuracy of the different alignment tools and pick a single tool that performs well with the type of sequences you have (long/short, protein/dna, few/numerous). Personally I prefer Clustal Omega. Once you know that you are producing quality MSAs then you can start to address the results of your AlignGVGD analysis.