Yes, this is correct if when comparing magnitudes with non-parametric tests. Unconstrained+flatten, this produces strictly positive values, there is no problem of flow direction. However, the tutorial about differences of source maps explains why this is not going to give correctly you all the differences between your two conditions.
I computed the difference of means (absolute value of average), put that one file in process2, and also put the one stat source map, then test>apply stats, but then I get an error "output argument "outputfiles" (and maybe others) not assigned during call to "process_extract_pthrest2>Run".
This process "difference of means" does not compute any significance level associated with your data. It is just a difference of the average, there is no statistical threshold that you can apply to this data. Use a non-parametric test instead.
As a general recommendation, it is better to start following the tutorials using the example dataset instead of your own data. This way you can focus on understanding the tools, and make you understand all the manipulations before reproducing it on your own data.