Which SCS method?
Are you using the same SCS loss, SCS unit hydrograph, and the same SCS storm? I haven't seen a problem running the same inputs in TR20 or HMS. What are your inputs?
20% sounds okay for different hydrologists who are making their own assumptions. I am a little concerned though that using 2 computer programs for the same methodology and procedures (I assume same model parameters) would give such a different answer.