The way I understand it, the two don't necessarily correlate.
Dual-channel, or more accurately, memory interleaving, pertains to the memory addresses, with the controller treating different banks as a unified, contiguous bucket, not unlike the virtual volume of a Fusion drive. It is a characteristic independent of speed, and not a new trick.
In the OP's first example, splitting the matched Pairs 1 (FE) and Pair 2 (FJ) between channels A and B may have enabled interleaving, albeit at reduced speed.
In the second example, keeping Pair 1 in Channel A, and Pair 2 in Channel B allows them to both run at full speed, but doesn't necessarily indicate that interleaving is in effect, since the controller is seeing non-matched DIMMs between the channels, and not group them.