for
1. the author only considered that person with lower ability would have more guessing behavior than other conditions. That's why the author invented the modified 3-PLM to simulate "guessing" response. We can visualize it from figure 2. The development of tailored testing and related methods may be only appropriate to Rasch model due to its statistically invariant parameter estimates. That's what the author explained in the figure 1. By the way, the 3PLM is notorious in parameter estimation (i.e., it is hard to estimate c parameter accurately). I think the modified 3PLM would suffer the same difficulty.
2. require more Monte Carlo simulation study? That is not considered in this study.
3. the author adopted y=15 for simulation condition. I think it's hard to decide the y value in empirical analysis. And the data may be insufficient to estimate the value of y.