Several years ago, I completed extensive transient modeling of a 150 CL liquid propane loading system. Due to the nature of liquid propane, the system contained several TSVs and operated with a small margin between operating pressure and line class in hot ambient conditions. Similar to your case, most of the system experienced surge pressures in the model that were acceptable, however, on the 3/4"x1" TSV inlet lines, pressures were found to be well above the line rating in the model for certain transient scenarios.
I discussed the phenomenon with AFT several times and the discussions were inconclusive as to whether or not the surge pressures were "real" or just an anomaly in the model. The uncertainty was due to transient cavitation leading to modeled vapour being > 10 % of line volume in the TSV inlet lines which causes "non-physical peaks" in the model and can lead to unreliable results. As a result, we conducted field tests and determined that the surge pressures on the lines in question were "real" and we measured this using "tell-tale" pressure gauges that indicated maximum pressures encountered at specific locations in the plant. At the time, the field testing correlated very well with the AFT Impulse model. I've also done other Impulse modeling that correlated well with field data.
As these surge pressures could indeed occur, we actually replaced the TSV inlet lines with 300 CL piping to accommodate the potential surge pressures. We also utilized temporary overpressure allowances from ASME B31.3 to accommodate rare surge scenarios in which surge pressures would be above line design pressure but within the 133% of design.
I do remember scouring the web for any references to similar occurrences but didn't find much. Made me and others wonder if this was an area of transient modeling that could use further research as from what I recall, I think the high surge pressures are related to the PSVs opening due to surge pressures and then slamming shut very quickly. Of course, in the model, I had to make some assumptions around opening/closing times which I took from a Tyco/Anderson Greenwood reference.
So, in my opinion, you definitely want to account for the TSVs in your modeling and hopefully, some more research is done on this topic.
Hope this helps.