This is my first post, and I would like to start by thanking ChemAxon for providing such a pleasant place to ask questions and report issues.
We have been using your tautomer generation on a large collection of drug-like molecules, and have been very satisfied with the results. There are, however, a few rare and notable exceptions. I will list two examples in this post. They have both been generated with the following options, with version 5.1.3_2:
cxcalc input.smi tautomers -D -a
The appearance of an alkadiene group was a little surprising and does not appear to be stable.
In this case, an amide group has been converted to an aminomethanol, which seems unusual, and also introduces a fairly significant structure change in the compound. In addition, if the output above is given to cxcalc as input, none of the other associated tautomeric states are produced, which indicates some type of inconsistency.