I have noticed what appears to be a problem in the ATR indicator. It seems that the AverageTrueRange indicator is not averaging the true range values correctly. When calculating the ATR for a period of 1 the values are what you would expect. However, when using a period of 2 the result values deviate significantly from a 2 period average of the true range.
To test this I created my own implementation of the ATR indicator and ran it in a system with the RE version. Both indicators where instanced twice, once with a period of 1 (the base line), and once with a period of 2 (the test). The system simply wrote out the values to a log file.
I then imported the first few rows of the data into an excel spreadsheet and added some manual calculations which illustrate the issue. Both the data file and the spreadsheet are included in the attached zip file. I can also provide the test system and ATR code for debugging / comparison if needed.