When reading in a thrust history file, the integrator sometimes misses a portion of the data in the input file.
For example, see regression test, THF_ThrustMass_CSC_None_None, this test fails because the integrator misses the mass flow rate occurring from t=3600 to t=3780 seconds. (This test uses the "ModelThrustAndMassRate" keyword).
However, regression test THF_ThrustMass_CSC_None_None_2 passes since the script breaks up the propagation into two segments. (segment 1 goes from 0 to 3600 seconds and segment 2 goes from 3600 seconds to 3780 seconds).
This case deals with mass but DJC thinks he has seen similar issues for thrust/acceleration data that has been missed.
A possible solution would be for GMAT to look at the times of the data in the input file and make sure that all the times are integration nodes "under the hood." (i.e., stop and re-start the integration at every time that is contained in the input file)
Setting MaxStep to one second and tightening the error tolerance to 1e-15 did not help.
Build used is GMAT-R2015a-TSF_Solve_For_Build-x86.zip delivered 9/22/2016.