The beam position on ETM looks quite low to me, and that means the beam is low on the primary optic of the TMS telescope too.

The only way to fix this properly is to give correct offsets to green QPDs.

Yuta calculated the QPD PIT or YAW normalized output against pure translation of the beam on the sled:

QPD1 output VS input: +0.733 /mm

QPD2 output VS input: -0.014 /mm

You need to divide these by 20 to get the offset that is required to move the beam by 1mm:

QPD1 offset VS ETM beam pos: +0.036/mm

QPD2 offset VS ETM beam pos: -0.007/mm

I started changing the offsets that Stefan set some time ago (QPD PIT = -0.464, QPD2 PIT = -0.084), and found that we need to go positive for QPD1 to move the beam up.  I didn't have time to center the beam on ETM and confirm that dither alignment didn't run away, right now the offsets are kind of arbitrary (QPD1 P = 0.232, QPD2 P =0), if you want you can go back to the original numbers.

Once ISI performance is better we should be able to dither ETM and ITM and slowly perform a2l decoupling using QPD offset (for ETM beam pos) and ITM angle (for ITM beam pos) as our knobs while Stefan's dither alignment is acting on ETM angle and one of the PZTs.