I am afraid I do not quite understand this paragraph. When do you get the same raster plot and when do you not?
I did not analyze your code in detail, but I think the way you translate the pre-synaptic spike time is still an issue. There is only one tSpikePre for the post-synaptic neuron, so if inputs are coming in from multiple synapses, they will not integrate correctly. Actually, I think the best approach for your synapse model would be to not include the int(timestep...) term in the equations, but instead have a variable that is zero or one (or actually a value > 1 if there are several incoming synapses that fire at the same time). You’d then update this variable from the synapse with an “up” and a “down” pathway, where the latter has a 2dt delay in your case. See this code here: Winner Take All mechanism using lateral inhibition - #3 by mstimberg
Hope this helps, let me know if this approach is not entirely clear.