fkine and ik_lm work wrong for fetch robot?

I want to calculate param of fetch robot using rt party.
My test code is below，it is very simple and the same method work right on panda robot:

import roboticstoolbox as rtb
import numpy as np

robot = rtb.models.Fetch()

init_q=np.array([0,0,0,0, -0.3, 0, -2.2, 0, 2, 0.7854])
start_pos = robot.fkine(init_q)
print("start_pos:\n",start_pos)

q_ik=robot.ik_LM(start_pos,q0=init_q)

print("q_ik:\n",q_ik)

end_pos = robot.fkine(q_ik[0])

print("end_pos:\n",end_pos)``

The picture of result like this

Obviously,the result is wrong.
I dont find the reason. Can anyone help me?

Start by running the supplied Fetch example and compare the printed start_pos and end_pos transforms. Read the Fetch model entry point and the fkine and ik_LM implementations, then inspect how the 10-element init_q is interpreted. Done means inverse kinematics returns a configuration whose fkine result matches start_pos, as it does for the Panda example.