Verification of pose estimation result via different approach

Days later I am back here to give my feedback and ask the way the result verified is correct or suitable or not. I have tried computing R and t with solvePnP but the result seems not that accurate according to the RMS which is or so with the 3D and 2D points for solvePnP. See my prior post here Way to verify or validate R and t given by solvePnP - #8 by FilipBaas

BTW, the depth info is from Kinect actually.

Then, I use g2o to get R and t in an optimization way. The result got in this way seems better. The RMS is about with the same set 3D and 2D points for solvePnP. This is what I want to ask. I mean is it Ok to verify the result given by g2o in the same way as done by the previous one. If it is yes, is this a result acceptable and what it is like in general solutions.
Thanks very much.

Guys, any suggestion would be appreciated.