Dear Guangpeng,
The bias potential can be shifted by any constant value without affecting the simulation (the forces remain the same) or the reweighting (it only adds a global shift to the FES). So, the fact that it is positive or negative is not a problem.
Regarding the original question, reweighting of metadynamics is a sensitive issue and several different methods have been proposed. A possible problem when using rbias for reweighting, is to correctly choose the length of the initial non-adiabatic equilibration part of the simulation to be discarded.
Best,
Michele