if all you want is to look at the timeseries, then yes, as long as there is suffiecent time to let the response fall after each block, you should be able to use deconvolution to estimate the average response to each condition. in your stim_file, you would put a one at the onset of the block for the condition of interest and 0's elsewhere, and estimate some number of TR's (lags) that will extend through the length of your block, plus a few extra to allow for return to baseline.