I did a graph of the Edison data and I found two big jumps for Biden - one was between 1-2 and the other was around 6:30.
maybe i am reading my data wrong or we are looking at two different data sets. i am looking at nyt_ts csv. are you unrolling the date field into raw minutes starting from wednesday 00:00 by hand, or relying on and trusting your graphing tool to do it properly for you?