Just speculating, but perhaps when on the ground & during climb, the optimum altitude on the cruise page is a 'predicted' value, based on the FMC calculated burn during climb & therefore the predicted TOC weight. And then, after TOC, it goes back to real time calculations. It would make sense for it to be that way.
That being said, I have noticed that the predicted crossing altitudes for waypoints during climb can be out by a few thousand feet, although this may have more to do with changes in track miles, temperature & wind during the climb.