First: for the application in the uBITX the 2 dB dip is no problem. But if you want to try and improve, change the number of turns (up or down) on L5 AND L7 and try tuning the capacitors again. If the dip gets worse, try changing L5 and L7 the other way. This action changes the matching to the filter to optimize it. I have no idea if adding a capacitor in the ground lead is a good idea, but I am almost certain that the filter unit should be well grounded and adding a capacitor will give a path around the filter and ruin the response.
Second: in your application in the uBITX, you don't need the improved steepness that using two units will achieve. I believe the original application for using two units to create a 4 pole filter is in two way FM radios at VHF and UHF frequencies. In this application the radio needs the extra filter performance because this filter is the ONLY filtering and it is setting the ultimate bandwidth of the radio to reject adjacent channel interference. In the uBITX, this filter at 45 MHz is a roofing filter and the ultimate bandwidth filtering happens further along in the signal path at the crystal filter. From reviewing a few data sheets, it seems that when intended for 4 pole filters, they are sold as matched sets, with a marking as noted above for the pins to be connected together in the middle of the filter. Shipping them in matched pairs suggests that they are somehow matched at the factory; I would guess they are matched to be precisely on the same frequency, though the polarity marking suggests to me that something more complex may be involved.
