1- after getting structure refined with phenix.refine to 0.145 , 0.164 Rw/Rf. and then you feel happy with that.

then after using validation server from PDB you get 0.158 , 0.181 Rw/Rf. why this big jump for R free. why phenix.refine giving us low R free where in reality is not?

I would guess something happened to the model file.
Can you reproduce R-factors yourself? Take data file and model that you sent to PDB, and use either of availbale tools in Phenix to compute R-factors:

phenix.model_vs_data model.pdb data.mtz
or
phenix.refine model.pdb data.mtz strategy=none main.number_of_macro_cycles=1

Do you get the same R-factors?

2- if i want to show kick map for metal binding site for publication, which map and how could i generate it.

Use FEM instead:

phenix.fem model.pdb data.mtz

There is option in the GUI too. More information
http://phenix-online.org/presentations/fem_06MAY2014.pdf

Pavel