Sir Elton . . . . .
I‘se . . . . . b a a a a a ck
Some reason or another, seems like that there is not enough inductance or not enough capacitance in the IF transformer(s) to pull the IF resonance on down to 455.
The most critical resonant pair of the TWO I.F. transformers available four sets of “ winding and adjustable trimmer capacitor “ involved, is the one that is off the plate of the mixer. The tuned sets trailing, then just enhance the overall amplitude of I.F. signal at that frequency, with broadness of tuning diminishing on down the alignment line.
Lets crunch some numbers in an evaluation, considering that the winding of that inductor/variable capacitor of that set is going to be a fixed constant , let’s use 1020 microhenry as its value. If one then sets a mica trimmer capacitor to being compressed up to ~ ¾ of its attainable max capacitance value, that will throw in ~120pf of capacitance and effect a resonance of the pair of ~ 455 Kc.
Now that is ideally, what we are looking for, but what you are EXPERIENCING, is 670 Kc, so that would be reflective of that existing L-C resonant circuit having only a capacitive value of 55 pf, present to work with that fixed 1020uh winding.
Now another aspect to consider would be the variance of the inductance factor, but that would require a dropping of the inductance on down to a reduced value of a whopping ~476 uh in order to create that tuning resonance of 670 Kc.
NOT gonna hoppen !
NowI can't eyeball the wire inside of that/those I.F. transformers , to see whether the wire beig used is common 35-38 ga single or double coton covered enamelled wire OR being Litz wire.
If its Litz, and taking 10/46 as the norm . . . and that would consist of 10 separate strands of #46 enamelled wire clustered together and loosely twisted as a single wire and then overwound with its cotton insulator(s) covering.
What I have seen happen on LITZ wire is the time related action of a chemical contaminant in eroding away individual strands , at a solder junction, thereby decreasing the effective inductance of the winding.
This was most common back at prodiction time in the erroneous cross utilization of a soldering iron that STILL had contamination of ACID flux on it from a counter chassis operation in mechanical soldering .
As individual conductors are eaten out of circuit , the effective inductance of the wound coil drops and frequency rises.
As a fix, one rosin fluxes and resolders the ends of the LITZ wire to get all conductors in circuit again.
BUT THIS could not account for the whopping 500+ uh involved here. . . . . only a 10 uh or so, if only hanging on by one strand.
About the only way to evaluate for LITZ presence in the coil winding is to peel back enough cotton overwrap, aside, as to then be able to see if there is a single wire or MULTIPLE fine wires.
Now considering that the fault is with low trimmer capacitance, how about working with the 1st IF transformer and padding in ~two ~47 / 50 pf fixed silver mica units . . . even ceramics in a pinch . . .to
see if the unit will tune in closer to 455Kc.
Also take that tip of disabling the local oscillator while aligning.
If this works, then the question is to be figured out as to the loss of available total capacitance of the trimmers. I couldn't even see severe salt water corrosion / rust ? deposits on ALL of the individual leaves of a trimmer, in limiting its max available capacitance down that much .
Work on that . . . . .
73's de Edd
Elton 
Edd 