What is the relationship between the number of state values and numerical solutions?

If the number of state values is high, the numerical solution (result) should be good.

However, even if the numerical solution is far from the optimal solution, the number of state values is high.

Does this mean there is a learning problem?

What's the reason?

Is critic mistakenly estimating the number of state values high?

What's the reason?　Is critic mistakenly estimating the number of state values high?

The first figure is the numerical solution, and the second figure is the number of state values.