So, let's look at some data that I've collected over the last 4 tanks of gas, starting from the most recent first (this is a AWD HiHy):

Manual (mpg) / Energy Monitor (mpg) / %Difference
-------------------------------------------------------------
26.9 / 27.7 / 3.0%
26.2 / 27.5 / 5.0%
24.8 / 26.1 / 5.2%
25.2 / 27.1 / 7.5%

These tanks were filled up at different gas stations and different pumps. Each tank had different driving conditions (A/C, %Freeway, Gunning, etc). The most recent tanks had the least % of freeway driving, and is interestingly correlated to a lower %Difference. However, there is insufficient data to claim that the correlation is statistically significant.

As you can see all 4 tanks show that the EM is consistently higher (~5.1%) than the manual calculations. This data shows good evidence that the EM reports systematically higher mpg estimates than that calculated by the manual method (total miles/gallons of gas). The NPR report provided from the previous post indicated that they measured discrepancies of typically 1/2 cup per 5 gallon or about 0.6% discrepancy, which is far lower than the 5% discrepancies from these measurements.

I find it unlikely that the different gas stations represented by these measurements are all systematically calibrated to report 5% more gallons than actually delivered. (Although it would make an interesting conspiracy theory.) I don't see how that could go unnoticed over time...

I'll keep monitoring this to keep adding to the statistics.