Comparison of the scale-independent values ⟨E_G⟩ determined for each LRG redshift slice correlated with the KiDS-1000 source sample, varying the galaxy-galaxy lensing analysis assumptions, sample and methodology. Our fiducial analysis adopted a point-based photo-z correction with z_B >  z_l, R₀ = 2.0 h⁻¹ Mpc and a model fitted to (Υ_gm, Υ_gg) including non-linear galaxy bias. The upper-left panel compares determinations of ⟨E_G⟩ varying the photo-z dilution correction method, the upper-right panel varies the small-scale parameter R₀, the lower-left panel varies the lens sample to consider BOSS and 2dFLenS separately, and the lower-right panel alters the fitting method to only include linear galaxy bias, and to use a direct fit to the E_G(R) values. The model line in each case is the prediction E_G(z) = Ω_m/f(z), where Ω_m = 0.3.