When the parameters file specifies downsampling by a factor of 16 when computing LFP, the low.kwd file does indeed contain a trace that is downsampled by that factor (that's good). However, it reports /recordings/0/downsample_factor = 1.0, and /recordings/0/sample_rate=25000 (which is the original sampling rate, not the downsampled rate of 25000/16=1562.5Hz). I'm not sure if the intended behavior is for these values to be 16.0 and 25000 or 16.0 and 1562.5 or maybe even 1.0 and 1562.5 (any of these would technically work), but with the given values it is impossible to reconstruct the actual time samples without having the parameters file around.
When the parameters file specifies downsampling by a factor of 16 when computing LFP, the low.kwd file does indeed contain a trace that is downsampled by that factor (that's good). However, it reports /recordings/0/downsample_factor = 1.0, and /recordings/0/sample_rate=25000 (which is the original sampling rate, not the downsampled rate of 25000/16=1562.5Hz). I'm not sure if the intended behavior is for these values to be 16.0 and 25000 or 16.0 and 1562.5 or maybe even 1.0 and 1562.5 (any of these would technically work), but with the given values it is impossible to reconstruct the actual time samples without having the parameters file around.