Linear on log graph paper.
It is hard to design analog circuitry that will take a log over 6 orders of magnitude (20 stops is 2^20 is about 1,000,000). The Pentax spot tends to have a long curve in them. Some worse than others. Calibration can reduce, but not eliminate. No doubt this is true of many meters. Some newer meters are very linear because the log is done with a microprocessor and the photonic reading is done with a very linear sensor. Ed is right though, there is no way to get your exposure to 1/10 of a stop. Too many other variables (film response, film speed, meter/film spectral response, shutter speeds, f-numbers)
So, yes, linearity can be an issue and the way to test is with ND filters. Over a short range, linearity will be pretty good, but over 10 stops you might see some variation. If you are worried about it, give it another stop or bracket.