I have not gone through the entire code (though you have a typo before the final Return in that there is not a semicolon after the integration before it--Returns in Mathematica are not necessary, but the original person who coded this was thinking as though he was programming in Fortran or Basic or C).

In going through the (typo-fixed) code line by line I find that the errors are generated in the final Do loop in the MT function execution. It is trying to take parts of AR that are too deep. Perhaps you have not specified your parameter definitions in your example case correctly? Beyond this, it'd require reading the paper in detail to understand things...