Here’s a summary:

1. The pointing information in eis fits header is for LW CCD only.

Note that the heliocentric coordinates stored in the fits headers and returned 
by the EIS software apply to the long wavelength band, not the short wavelength band. 
This is because the He II 256 line was used to co-align with the SOT data.

2. Images on SW CCD are 16-20" (or pixels) higher (ie. in Y-direction towards north pole) than LW CCD.

This may vary with wavelength

3. Images on SW CCD are 2" (or pixels) righter (ie. in X-direction towards solar-west limb) than LW CCD.

The fact is that the two CCD images arise from different halves of the primary mirror. 
The focal points of the two mirror halves could thus be different.

4. Slit-tilt effection?