Look at the N for high GPNMB (>=25%) IC group, only 8 subjects, 3 subjects in triple negative and high GPNMB IC group, so most likely N would be much less than 4 for ultra high GPNMB (>=50%) IC group, probably 0 for triple negative and ultra high GPNMB, the sample is just too small to release as another separate group.
Agree it is undoubtedly a small group - but ultra high G was one of the predefined subgroups and probably about as large as the non-predefined subgroup that they did release (HG/TN). My guess is that, unlike HG/TN, ultra high G does NOT show meaningfully increased efficacy over high G (which was almost entirely driven by TN/HG subgroup anyway).