In the half-step part of the analysis, we walk through the update period and play back all of the changes, hour by hour, day by day. From the two XER files, we have actual dates, remaining duration at the start and end of the period, etc., so we can measure the impact of individual changes on the finish date.
The revision half, which includes logic changes, is another story. We have no data on when the logic changes and other revisions were made, so we cannot know the order in which they were made. It turns out that the order in which the revision changes are applied matters very much, so we are unable to calculate correctly the impact of individual changes.
We attempt to give you a clue by showing you the fb and fa columns. The fb (float before) is the float after the half-step but immediately before the revisions are applied. The fa (float after) is the float value after the revisions are applied. So, for example, if an unstarted activity has 5 days trimmed from its remaining duration and float before and after is 0, it's likely that the finish date had some recovery. And if the float value before and after is a large number, it probably did not impact the finish date. But we cannot quantify it and be correct mathematically. At least, we haven't figured out how to do that yet; we haven't given up.