Just a note to be very careful about reading too much into a synthetic spread expression (the Lavender bars on the chart below). This is especially true if one or more of the spread legs has less frequent trade prints than another. The net effect will be an exaggerated trading range for individual bars - short and long time frames. Furthermore, synthetic spread expressions will give a false volatility reading as well. Use the exchange spreads for the accurate (the Blue bars in the chart below) and true spread pricing and spread trade volume. You will need to check with your charting platform provider to find out if they 1. Indeed provide exchange spread data, and 2. Where to find their index of exchange spread expressions. Unfortunately, the symbols used for the same products and spread expressions are not uniform between exchanges and data vendors (this has always really pissed me off btw).
It's interesting to me that, after not really being mentioned in my memory, there have been a good number of "reality-not-comporting-with-model/theory/precepts/...." comments in ET posts lately. ("For the last 2-3 weeks or so??") You post provides solid evidence.
ESignal pro-tip: if you have to use fully synthetic spreads (i.e. Cattle crush or other inter-instrument stuff) you can use a 1440 timeframe for daily bars (and it will use intraday data) or if you have 12.7+ you can use things like 7D rather than W. The only reason the exchange traded spreads are better for the typical condor, fly, double fly stuff is because the calendars get the most liquidity by far. Try doing it with exchange traded flies or condors and you'll see it sucks for anything other than Eurodollars (which get a shit ton of liquidity on basically everything).