Having abandoned the idea of using the GIZA++ alignment scores, I can now switch back to the more reliable Moses alignments (a smart combination of both GIZA++ alignment directions). Having done so, the BLEU score jumped about 3 points:

r157-20080731053726: BLEU = 20.05, 52.9/27.1/16.8/11.0 (BP=0.885, ratio=0.891, hyp_len=4809, ref_len=5398)
r169-20080805193448: BLEU = 23.21, 52.3/28.2/18.3/12.4 (BP=0.966, ratio=0.967, hyp_len=5219, ref_len=5398)



Interestingly, the score with Moses in inclusive mode dipped by a point, presumably indicative that the statistical models are more accepting of our translations using the improved alignments:

r157-20080731075014: BLEU = 30.78, 58.2/36.1/25.6/18.6 (BP=0.974, ratio=0.974, hyp_len=5260, ref_len=5398)
r169-20080806022831: BLEU = 29.92, 57.0/34.7/24.0/16.9 (BP=0.999, ratio=0.999, hyp_len=5394, ref_len=5398)