What I thought was going to be a simple search:
lucy | eval UsedT=UsedMB/1024/1024 | eval UsedTB=round(UsedT,1) | timechart avg(UsedTB) as TBUsed span=1w | predict TBUsed
But my graph is showing the current data, up to today (March 28th), then jumping back to February 1st.
Root cause was found. Turns out everything gets screwed up if there is missing data. In this case my data set was only about 90 days, and I was search for 120 days. The null data at the beginning screwed everything up.