I also tried other JFETs, a 2N5458 with a 180Ω. or MPF102 with a 220Ω, source
resistor, and got results pretty much the same. I kept the 2N3822, as its data sheet states that it's for VHF work, and it seems just noticeably better.
And here is a method of checking your junk box JFETs.
FM fool's predicted NYC dial:
One version of the new FM front end uses a VHF TV set balun. But the JFET will see everything in the VHF spectrum. Another version uses an LC circuit tuned to around 99MHz. One turn of insulated wire forms the input from the FM antenna terminals but isolates the antenna terminals from the hot chassis shock hazard. Only thing here is that the radio will be less sensitive at both ends of the FM dial. Using a pair of LC circuits (one around 91MHz, the other around 105MHz) coupled via induction and stray capacitance to give a reasonably flat band pass over most of the FM dial helps here. Using a grid dip meter helps a lot in getting these LC circuits on these frequencies. I used a single ferrite adjustment slug mostly for the 91MHz LC circuit, I had to fiddle with trim caps to get both resonances where I wanted them. The single ferrite slug helps couple RF antenna energy across the LC circuits and into the RF amp stage. You can check the grid dip meter's calibration by seeing what its dial says when you jam an FM station of known frequency.
Closeup of the FM front end's input and band pass filter. The LC circuit further from the circuit board is peaked around 105MHz, the closer one 91MHz. Antenna coupling is an almost single turn of wire fed from the antenna terminals. Maybe this makes it around 300 ohms? Probably too short to matter. Reception is slightly better than that I had with the input balun. The curve is just a guess, stations around 98MHz come in pretty well, so they are likely just "couple o' dB" down. Sure, it might be better if I had a 3rd FM tuning gang on the tuning cap, but I don 't... I did try a tuned circuit on the FM RF amp's input (built it with a grid dip meter to get its frequency range correct) with an independent tuning cap, but it didn't seem effective, about a db at best. Not worth the trouble of mechanically rigging it to the tuning dial.
The modified radio, with replacement resistors and electrolytics. And note the white wires with orange and blue (Syracuse!) colors on the tuning cap.
Inside this homebrew set is the RFI filter, and series resistor.
This RFI filter knocks down crud from the powerline. Note, right of the power resistor, the white wire with orange and blue, like Syracuse Orange and Blue.
If you are working on GE radios like this, check the carbon comp resistors. I had to replace a 100 Ω, a few 150 Ω and a 220 Ω resistors. They all went flakey. The bad 100 Ω resistor made the FM local oscillator quit at the lower end of the band. In a few spots, where the circuit board nodes were close together, found it easier to use the larger sized surface mount resistors to replace the old resistors that were in crowded areas on the component side of the board. While you're at it, check the rest of the carbon comp resistors, most in my radio went high by 20 to 60%. I decided to change them all out. After replacing them, the radio got more sensitive, like I remember it when it was nearly new. The resistors in the green RC circuit modules (some examples below right) have held up very well, not drifting in value. No need to change those out. These in the picture came from other GE radio circuit boards. My father used to get these boards from some surplus shop on New York City's Canal street back around 1970.
Above left: unusual resistors, essentially single ended. Likely to sub for regular resistors that would have one lead folded over to have them stand up vertically on a crowded circuit board. They look like the sort you find in Japanese transistor radios, except these have both leads at one end. Usually, to mount an ordinary resistor vertically, you have to fold over one of the leads to run parallel to the resistor body to get back to the circuit board. The pictured resistors look to be a reasonable answer to this, don't know why these didn't become an industry standard. Look to be designed for use in crowded circuit boards. That pesky lead on the top is not exposed. These resistors look to be made using hollow ceramic tubes with resistive material on the outside (under the paint), and one of the leads passing through the hollow tube center and terminating at the top, and the other lead terminated at the bottom dogbone resistor style. Then they were dipped to coat them with the paint and then markings. Used a few of these in the above radio.
Above right is the cabinet my radio circuit board was intended to live inside.
Speaking of resistors, they made 1/8 watt carbon comps. Go easy with the soldering iron or dissipation on these, else the heat will cause the resistance to go up, making them go out of tolerance.