Fake data, estimated data, or diagnostic data?

After reading the numerous accounts of unexplained quantum leaps in Cox data assessments, I have become diligent and regularly check my daily and weekly metered usage on Cox.com.  Over the past year, the Cox meter says I typically use just under 300GB each month and the weekly pattern is consistently uniform.  Refer to Figure 1 and Figure 2 in the linked JPG image:

www.dropbox.com/.../data use meter.jpg.  

These are screenshots of my data usage graphs copied from my account information on the Cox web site.  The weekly pattern corresponds to my usual internet habits and the claimed data consumption is similar to a rough estimate.  So I think Cox claims regarding the accuracy of the data meter do have merit.

This Fall and Winter, however, there were some strange anomalies, though I still have not come anywhere near the 1TB cap.  During November, a windstorm toppled a tree that severed the power line to my entire street.  We lost electrical service for 36 hours while the affected pole and wires were replaced by the utility.  Naturally, I expected the data consumption meter to show at least one day with zero use.  But the graph for that week was virtually identical to the weeks before and after the incident, which was curious.  Perhaps, I theorized, Cox is able to power the phone modem and cable modem by supping DC over the cable, similar to my outdoor antenna amplifier, so data continues to flow despite lack of electrical service. 

Between Christmas and New Years we traveled completely across the country to spend the holiday with relatives.  Since no one would be at home to use the internet, I not only switched off power to the modem and router, but also completely disconnected the incoming cable (for 10 days) while we were away.  What resulted was completely the opposite of what I expected; according to the Cox data meter, daily data consumption during those 10 days was much higher than normal, and did not match the usual weekly pattern.  Once the cable connection was restored on January 2 , the historic consumption pattern magically returned.  Refer to Figure 3 in the linked JPG image for a labeled screenshot.  

Had I left town for a month instead of 10 days, I would have exceeded the data cap without actually using a single Byte of data..

So, if I accept the Cox data meter is accurate, how were these 2 events possible?  I have three theories that may explain both, as well as some of the experiences reported by others:

Estimated data
When a day occurs with zero data consumption, the Cox meter (which is probably an outsourced computer program) algorithm assumes an error occurred and substitutes an estimated data consumption based on historical use.  Utility companies regularly substitute estimates when bad weather or employee absence impact the meter-reading schedule.  An estimated water bill that seems too high or low does not mean the water meter is inaccurate.  This theory could explain why the day-by-day data consumption did not change during the week we lost electrical power.

Fake data
When the cable was disconnected, the Cox meter algorithm interpreted this as an attempt to defraud the company, so invoked punitive countermeasures, one of which is to assume you are actually using data at a high daily rate, but somehow blocking the meter from seeing it.  The same thing could happen if a external system failure interrupted the cable connection to a residence or multiple residences.  Often, when the electric utility thinks a customer has tampered with the meter, they over bill to compensate, based on some assumed actual use.  It becomes the responsibility of the customer to prove innocence and challenge the "fake" electricity they have been charged for using.  This theory would explain why the daily data usage during the aforementioned 10 days was fairly uniform, rather than following the
historic pattern of higher use on weekends.

Diagnostic data
Perhaps, when the incoming cable was disconnected and/or the power to the modem was off, Cox began to transmit  diagnostic data in an attempt to determine if there was a regional failure.  Evidently, the data meter does not differentiate between diagnostic and requested data, both flow from Cox to the customer, and are consequently recorded as data use.  In other words, the Cox meter measures the data Cox sends your way, not what your modem actually receives.  This might explain why some folks who independently measure inbound + outbound data with their own modem find it does not match the consumption reported by Cox.

I am not convinced that malice was intended.   More, likely, no one at Cox knows how the data meter works and what policies it enforces, because it is outsourced.  This would explain the often-silly advice offered by customer service representatives when someone questions an unexplained spike in claimed data use.