HCP Task Timing, Duration, and TR Discrepancy?

I read the Human Connectome Project S1200 release manual and its Appendix VI for information on the fMRI task data, and I am noticing a seeming discrepancy between the number of TRs (frames), the length of the TR, and the “Run Duration.”

For instance, the TR for all the fMRI data is 0.72 seconds. In the Emotion Processing task, it says that the number of frames should be 176 (which is accurate to the data). This means that the total run duration should be 0.72*176 = 126.72 seconds, or 2:06 min:sec. However, the manual says that this task has a run duration of 2:16 min:sec, which is off by exactly 10 seconds.

Similarly, all other task scans are off by exactly 10 seconds between their computed run duration with TRs and the reported run duration. Is there any information provided to explain this discrepancy? Are there any forums that discuss this?

Next, it is unclear to me what time in the EPRIME/*TAB.txt files correspond to the first TR of the scan. In Appendix VI in the manual, it states:

SyncSlide.OnsetTime - onset time of countdown slides before task; the first row in the SyncSlide.OnsetTime column reflects the onset of the first TR in E-Prime clock; to ensure correct EV timing, that value should be subtracted from clock time for other trials.

The EVs/ files are indeed computed in this way, but then the times cannot be fully accommodated by the available TRs.

For instance, in the Emotion Processing task, the last fear trial in the last fear block should have a time of 131.3 seconds, and yet as I mentioned above, there are only 126.7 seconds worth of TRs.

How do the EVs and EPRIME/*TAB.txt files align with the TRs in the fMRI data?