Normally gaps are inserted as part of the alignment algorithm for clustalw, if you are looking for "ungapped alignment" the only way I can think of is setting a high -gapopen penalty.

I'm not sure if there is a tool that will suit specifically to your situation, but you can find useful links to various MSA tools in one of my older answers: A: Analysis of a conserved motif across a phylum

I am usually quite happy with the results from the T-Coffee multiple sequence alignment server: http://tcoffee.crg.cat/

Did not use it for this particular purpose though, but with the right parameter setting (e.g. using local aligners for the pairwise alignment step, running T-coffee in 'Advanced' mode setting the '-gapopen' and '-gapext' parameters to very high values, e.g. 1000) you should be able to do what you want.