Have you tried adding 1 to the binary indicators (so that they're 1/2 respectively)? Not sure if the same applies here, but LCA models that I've run in poLCA have issues with zeros in the dataset.