All my constants were lost if there were any other than the ones that the calibration in the Enhanced Accuracy menu stores. I suggest redoing that and checking your frequency measurement again.
I *sometimes* read 100 kHz within 5-10 Hz if I have reasonable settings, but small changes in parameters will result in errors of 200-300 Hz or larger. Hardware measurements are generally the least accurate for frequency and period with 1% errors. In general a sampling scope is not well suited to such measurements.
The 11801 FW is *extremely* buggy. I would expect that the CSA803 shares the same problems. Under certain conditions which I cannot reliably reproduce, I get rise time measurements of the calibrator step of 300+ ps. This is often accompanied by 3 glitches on the rising edge near the start, middle and end of the step. This will persist through a power cycle and can *only* be reliably cleared by using the "initialize" option in the Utility menu.
I'd only had the instrument a short time when I ran into this and was quite freaked out when I encountered it. It took a good bit of poking around in menus to sort it out.