> I assumed that the output I sent to you would refer to the TRs. Or is this wrong?

If you mean the output from 3dDeconvolve, then each sub-brick correspond one value for a specific regressor the user specified, which has nothing to do with TR.

> So what other approach for response function modeling would you suggest instead?

That depends on the experiment type, specific tasks/conditions, etc.. Even with deconvolve, nowadays you can use basis functions such as TENT instead of adoption option -stim_maxlag.

> Does this mean that I cannot use the obsolete approach at all for connectivity
> analysis, or is it ok if I delete the maxlag specification for SEED and interaction?

You can still use -stim_maxlag for hemodynamic response modeling if you want deconvolution, but for your seed time series and the interaction regressor there is no point adding -stim_maxlag.

Gang