[R-390] R-390A Cosmos PTO has erratic linearity

[Larry H]

I'm restoring a Stewart Warner R390A with a cosmos pto type 136-1. Although I've had to fix a number of problems on the R390A, it looks and operates well now except for 1 issue (or I think it's an issue). I've graphed the linearity of the pto numerous times every 25 KC, and it tracks mostly +/- 100 - 200 cycles (with a few exceptions up to 300 cycles), but after adjusting the linearity bad spots numerous times, different ones keep popping up.  IE, the measurement repeatability is poor, but still mostly with in +/- 300 cycles.  This is after letting it warm up for 2 hours.  The frequency stability is very good.  Also, if I graph the pto linearity, use it for a few minutes, and graph the linearity again (without making any adjustments to the pto) its different than the previous graph.  Everything inside the pto seems tight and the gear train to the readout is tight (IE: very little slack). I have cleaned the pto shaft grounding contact and checked
 the contact pressure.  The endpoint also varies about +/- 100 cycles. I have the prescribed spacing in the oldham coupler (its in very good condition) and it is ligned up very closely with the KC shaft and the spring is installed correctly. The gear train and the pto operate very smoothly.

Should this be better or is this normal?  Any suggestions would be most welcome.  Thanks.

Regards, Larry