In a private message, I got some information and encouragement to keep focusing on that Interface Products IP8300. One question was about the behavior of the LED lights. They are, indeed, not acting right

The LEDs as the car approaches the 3rd floor are "erratic" as follows.
* car slows
* LU comes on
* DZ comes on
* LU goes off
* LD goes on
* 1FP goes on - car stops with LD, DZ, and 1FP illuminated

I think it should go
* car slows
* LU comes on
* DZ comes on
* LU goes off
* 1FP and 2FP (1+2=3) both go on - car stops with DZ illuminated (and those floor indicators)

The current tape magnets are glued on really tight. I started to try to free the rightmost one to move it up, but it was likely that I was going to break it so I stopped trying that until I was more sure.

I thought I had found some info about the sensor device.at

http://www.jrtinc.com/webroot/uploads/ascenseur/J12-1000_A_07-3.1.pdf

That information almost matches this elevator's IP8300 sensor device. It does, at least, explain the principles of how it works.

I experimented moving a little test magnet around the face of the sensing unit to activate the IP8300 magnetic switches. That told me enough for me to realize that the diagram in that webpage is not identical to this elevator's IP8300. So, I opened up the IP8300 and documented the actual configuration of the magnetic switches.

So, my version of the IP8300 device has three vertical rows of magnet switch sensors -- most of them in the center row. Interestingly, the switches are sensitive to the north vs south pole of the magnet used for testing. {I don't know the polarity of my test magnet so the info about polarity below might be backwards.}

Starting from the top, there's a 4FP magnetic switch in the center, activates with "south". This is 1/4" from the top.

Next, 3 9/16" from top, the LU (level up) switch is in the center, activates with "north".

Next, 6 1/8" from top, the first of two DZ switches is in the center, activates with "north".

Next, 8 3/8" from top, there's no switch in the center. There's 1FP and 2FP to the left and right. (1FP is on the left side of the metal tape). These activate for "south".

Next, 10 3/8" from top, the 2nd of two DZ switches is in the center, activates with "north".

Finally, 13 1/16" from top, the LU (level up) switch. Activates with "north".

So, I put a temporary magnet on the 3rd floor above the existing magnets to "trip" the 4FP switch. I got excited because suddenly the elevator would work for the 3rd floor. With the elevator at DZ (door zone) and 4FP "on", the elevator stops at the third floor, unlocks the door, and displays #3 on the car panel display. In other words, the elevator works fine at the 3rd floor (with that magnet on the tape to trigger 4FP). It was not exactly level though - now the car was 1" high.

While testing, I found that it was not reliable. Most of the time, it worked. Sometimes, not. But if I moved the tape/magnets closer to the sensors, it would suddenly open the door lock. In that case, 1FP and 2FP were both "on". (Note: in binary, 1+2=3 for 3rd floor). That got me thinking that the magnets were weak or not close enough to the IP8300.

So I measured the distance of the metal tape from the shaft wall. At the 1st floor, it was 2 1/2". At the 2nd floor, it was 2 1/8". At the 3rd floor, it was 1 7/8". You would think that the nylon guides on the sensing unit would pull the tape to just the right distance, but apparently this is a place where 1/16" can really matter.

I decided to get the tape to be closer to the sensing unit and vertical. First, I drilled out the four screw holes on the nylon guides so the guide slot was 1/16" closer to the face of the IP8300 sensing unit. Next, I bent the support bracket for the metal tape (at the top of the metal tape, above the 3rd floor). I set the metal tape to 2 3/8" from the wall.

Now the elevator works correctly for the third floor. It levels up nicely too.

Conclusion: The metal tape was not hanging perfectly vertical. At the top floor, it pulled away from the sensing unit ever so slightly. Sometimes it would work, sometimes it would not.

I really appreciate your time, effort and knowledge of everyone's responses.