problem with the normals of reduction model

Hi,
The model can be decimated to a small number of faces, but the normals may be adjusted to get a better vision.
The original model file:https://app.box.com/s/2ycf7j6uc5lzlqtj4rj0acvhuhdvzwyl
The full resolution picture after join vertex and normals:https://app.box.com/s/9fx76l2r9bw4tbm8tw8i4nhg4tua4e37
The reduction result:https://app.box.com/s/7p73g8wf2sw2xdueuvi0mxx9kikjb35t

Of course, I can get a better vision to select a high resolution.
But I am just wondering is there a way(eg. adjust some params) to get a better vision with lower resolution.

Thanks