1. The discrimination variance implies nonuniform DIF.
2. Yes. It is odds that the authors believed the item property could be used to decide which item is DIF items.
3&4. I totally agree with the opinions.
5. Since there is no comparison between the proposed models and other methods, the studies is not convincing enough. But, I think the authors make great efforts to define the models according to the difiniton of latent DIF.