I am double checking to make sure my graph in reply #30 is correct. Is it correct?
Looks like it is.
I have read (RSGB Handbook) that leaving the output side of the link antenna coupler is best kept floating. Grounding this side of the coupler (center of tank coil or rotor of split stator capacitor) increases the potential of the Marconi effect. http://amfone.net/Amforum/index.php?topic=12469.5;wap2
Too right! Rob, according to my schematic of the 275 watt matchbox (not sure what size yours is, or if all sizes use exactly the same circuit), the coil is indeed floating, but the rotor of main split-stator capacitor C1 is grounded, and the two inner-most sections of differential capacitor C2 are tied together and grounded. That is effectively the same thing as grounding the mid-point of the coil. Try lifting the connection where the rotor of C1 connects to the sections at C2, from ground (which I assume to be the metal case of the Matchbox). I would also assume the ground terminal simply connects to the case. Leave the connection between C1 and C2.
The connection to C1/C2 and ground is necessary when the Matchbox is being used to feed a single-wire antenna, since that is the only way the circuit has a ground return. But when feeding a balanced two-wire line, the ground is unnecessary, and in fact, may cause unbalance at the feeders precisely for the same reason grounding the mid-point of the coil would: it is providing a ground return for common mode current and allowing the feedline/dipole combination to work like a vertical-Tee against ground in addition to working like a balanced parallel line fed dipole.
Since the common-mode current is of the same polarity in both feeders, but the differential current is of opposite polarity in each feeder, the sum of the currents will be different in each, since in one feeder I (total) = I (common mode) + I (differential), and in the other feeder, I (total)= I (common mode) − I (differential). In additional to the unbalance, the common mode current (probably much less than the normal differential current) nevertheless causes some feed line radiation, plus ground losses if the Matchbox is earthed to anything less than a good low-loss radial ground system.
Lifting the ground point at C1/C2 won't blow anything up, since this would not substantially change the voltages across the capacitor sections or at the band-switch connections. If anything, it might actually reduce them, since unwanted common-mode voltages/currents are not being added to to the desirable differential currents at one feeder. Lifting this ground will not affect the ground return at L1, the link. If you were using a link-coupled final and a balanced twisted-pair or parallel-line link instead of coax feed, there would be no need for this ground connection, either.
I would modify the Matchbox to have an external ground strap to C1/C2, that could be removed or added as needed. It shouldn't be needed with a balanced open-wire transmission line. It is incomprehensible to me why Johnson didn't do it that way to begin with. I would keep the external ground connection to the metal case of the Matchbox, to keep the unit effectively shielded, and to prevent the box from becoming hot with RF, in case something goes wrong to unbalance the system, such as a conductive object contacting one of the feeders, or one leg of the dipole coming down.