Previous kernel incorrectly modified both amplitude and phase:
(W*R/(1+W*R)) * exp(-j*S*R/(1+R))
Corrected to unit-amplitude pure-phase kernel as defined in the paper:
K = exp(j * S * W*R/(1+W*R))
Also updated default thresholds to symmetric ±0.35 which is appropriate
for the corrected phase range of approximately [-S, S] = [-0.48, 0.48].
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
