The schematic helps indeed. Both 74LS parts used in the schematic are "slow" when it comes to rise time. The JK filpflop, according to the datasheet I checked, has a typical max clock frequency input of 20MHz, so close to what is used in this design.
The 6MHz signal, like you stated in your original post, comes out as 33% or 66% pulse width depending on which output is measured. On the green marked one it would be high for two clock periods and low for one. This can be seen on your scope image. But there it has a rise time of ~50ns which is indeed quite long even for the 74LS107.
So maybe what others have pointed out is that you are using 1x setting on your probes.