A few questions about using the LDA+U method to get the right band gap

[xiaogai]

Dear users and admini.,
I intend to obtain the experimental band gap value by the LDA+U method. But I encounter a few puzzles.
1. Are the parameter U and J unique for a given material? In some paper, the U and J is chosen randomly only to reproduce the experimental band gap, and in other paper, the U is definitely determined with constrained-LDA calculation. Which method is proper?
2. Why for ZnO, the LDA+U method can not correct the band gap calculate by LDA to the experimental values.( ZnO(LDA)=0.7 eV, ZnO(LDA+U)=1.5 eV, ZnO(exp)=3.4 eV? Is it because of the choosen of parameter U and J? Why for TiO2 , the LDA+U method can well obtain the experimental band gap value?
3. Is it true that if proper parameter U and J are chosen, we are sure to get the experimental band gap value by LDA+U method?
4. When employing the LDA+U method, whether our calculations need to relax (ISIF=3) again if we employ the relaxed structure obtained without plus U?

Thanks a lot!