1. Time series regression de-trend and de-seasonalize time series data by using deterministic approach, i.e. dummy variables for each month, and assuming linear trend over time.
2. Box-Jenkins Methodology removes trend and seasonal components through differencing, which is in a stochastic fashion.

I would like to know which one you prefer when you are working on time series and which one your company prefer at workplace?  Off course, both methods have their advantages and disadvantages.  Theoretically the way to distinguish when to use regression and when to use Box-Jenkins was widely discussed, and tests such as Dickey-Fuller can be used to test whether the series contain stochastic trend.  However the problem is when you at work,  stakeholders want to see reports with consistent methodology and summary results in the same manner.

Also in R, time series analysis to me is hardly to be implemented. The main problem is for Box-Jenkins, I cannot find a package that can handle transfer function easily. If anyone knows about it, please let me know.  So recently most of the tasks I completed were in SAS. However I try to stick on one software for the analysis.