Thanks for the links to timenuts, I thought I was going crazy or there was something wrong with the unit. After posting this I did a deeper dive on the timing of the pulses with my HP 5371A. What I found is that the pulse width and spacing are within 70-100pS, 120pS worst case (over a 1000 pulse sample) regardless of the variances I am observing in pulse train frequency. After reading the "theory of operation" in the service manual I became suspicious that the unit was only intended to generate precise pulse widths, delays but not continuous pulse trains, this also explains why pulse train frequency accuracy improves with a wider pulse width. Reading the responses at timenuts verified that there was nothing wrong with this instrument, very interesting reading.
Best Regards,
Craig Petersen.