It would be easier to look at if the birth dates were represented: each of the vertical intervals in time has a cluster of outgoing edges at the top for descendants. This makes things look cluttered and loses information. If these outgoing edges came from the right distance down the interval for the birth year then it would be easier to see structure in the graph.
Even for this early version I had debated whether it was worth it to do it for descendant tree due to clutter issues (it can still happen for ancestor trees, but the heuristic limits used make this a little rarer). I'm not quite sure I understand what you're suggesting though - the curves coming into the top do terminate at the birth year, but the node itself kind of covers this up - is that what you're referring to?