Robot Calibration DH-Parameters mismatch

Hello, I am trying to extract the DH-Parameters from an e-series UR5e to get accurate inverse and forward kinematics in python.
I have extracted the parameters both from the calibration.conf file and from an unzipped .urp file.
When using the values from the calibration.conf file and adding them to the “standard” DH-values (from urcontrol.conf.UR5) I get quite large values for d and a. :

delta_a = [
-5.4451392440e-05, 
0.02412010075500, 
0.0012052715012, 
4.35400296829e-05, 
1.19843008576e-05, 0]

delta_d = [ 
0.00022243, 
-248.57141, 
252.824383, 
-4.25266815, 
-2.47625006e-05, 
-9.50598885e-05]

after adding the deltas these are the calibrated values:
image

As per this post these large values are to be expected, as they cancel each other (especially the 250 meters of d[1] and d[2].
However, there is one value that does not seem to cancel out:
The second delta_a-value (~24mm) does not seem correct and is not cancelled by any of the other delta_a values.

When creating robot targets via inverse kinematics with these values I get a discrepancy of ~50mm depending on joint configuration. My guess would be that the 24mm discrepancy of a[1] is to blame.

What am I missing here? Where does the 24mm delta come from and how can I account for it?