In a scientific correspondence to Nature in 1988, Steel, Hendy and Penny quantified the loss of information inherent in phylogenetic analyses based on genetic distances. In spite of that loss, distance-based methods for inferring both trees and networks have proven both useful and popular. Some even argue that they should work equally well for large data sets. Here we engage with the original debate, and explain how the loss of information in both distance and character data has stymied best efforts to construct effective recombination tests with NeighborNet or SpectroNet.