While meshopt compression (EXT_meshopt_compression) is generally used alongside quantization (KHR_mesh_quantization) for best compression and memory optimization, there's no technical requirement that they be used together. Currently glTF Transform just doesn't support that, it always does quantization when doing meshopt compression.

This issue tracks a feature request for allowing meshopt compression without quantization. Compressed sizes will be larger, but compatibility and quality will be somewhat higher.