Dennis,
You have an error in the free space velocity which Albert mentioned indirectly.
Multiple reflection in the cable produces one or more poles in the transfer function. The SG503 may compensate for reflections from the ends of a 3 ft cable. However, something as basic as physical robustness of the cable may be the primary factor. Just making wild guesses.
I'm not at all familiar with the SG503, but I have been doing a lot of TDR of cables and connectors with my 11801. I see a lot of not quite 50 ohm cables from reputable vendors. It's a bit disappointing. It also leaves the question of how close are the connectors to 50 ohms and over what BW.
I suspect that the cable with the reflection in the middle was partially crushed or kinked at some point.
I'd really like to see the no cable case at a faster sweep rate, say no more than 100 ps/div.
You can see some of what I've been doing here:
170 in particular shows the reflection from a BNC-F to BNC-M connection between two cables which otherwise match. I'm fairly certain they came from the same spool of cable. The two cables are BNC to SMA jumpers a friend made up for me when I was singing the "Chinese Connector Blues".
Have Fun!
Reg
