Problem in two phase initial solution search

In testing Ludovic’s model, I’ve discovered a bug in my two-phase initial solution search routine. Better take a close look at this.