Created attachment 2838 [details]
Simulation logs and the patch for the scenario

Hi, 

The bug is related to the lbt code where an LAA node starts transmitting during the ongoing TXOP of another node, in spite of sensing the other node above its ED threshold.   

#time(s) nodeId	type	sender	endTime(s)	duration(ms)	powerDbm
2.85	   3	lte	1	2.850214285	0.214285        -53.6777
2.85021    3	lte	1	2.850999999	0.785713	-53.6777
2.851	   3	lte	1	2.851214285	0.214285	-53.6777
2.8512	   3 	lte	3	2.851999999	0.733714	23
2.852	   3	lte	1	2.852214285	0.214285	-53.6777
2.852	   3	lte	3	2.852214285	0.214285	23
2.85221	   3	lte	1	2.852999999	0.785713	-53.6777

From the third row of above logs it can be seen that sender node 1 is only transmitting a CTRL signal of duration 0.214285 ms during one subframe (1 ms). Due to this, node 3 finds the channel free for rest of the duration of a subframe and starts transmitting its reservation signal and rest of its transmission.  

This problem was reported here: https://groups.google.com/forum/#!topic/ns-3-users/iVHnP20oLhY

I used the http://code.nsnam.org/laa/ns-3-lbt repo for the simulation.

All the simulation logs and necessary changing to the LAA simple scenario to reproduce this behavior are attached here. 

Kind regards,
Zoraze