If I see it correctly the monthly (weekly delta is calculated by subtracting the latest from the earliest data point in that month.
This seems obvious, but leaves out the delta between the earliest data point on the month and the latest data point from the previous month.
This leads to strange aggregation like:
Year to Date (Jan-May): -3.4kg
CHANGE REQUEST: subtract latest within the current week/month/year from the latest in the previous.