How much have you experimented with this? Here is what I see, tested on builds 2580 and 2660 using the wxMach screen set. You didn't mention which Mach4 build or which screen set you were using (which is useful information when reporting anomalous behavior).
(1) With default jog increment settings, I can cycle through 1.000 through 0.0001 step sizes as expected.
(2) Go to Configure->Mach... and change the "Jog Increments" table as follows:
2 0.2 0.02 0.002 0.0002 3 0.3 0.03 0.003 0.0003
then click on "OK" (or for the Mach3 paranoid, "Apply"and then "OK").
(3) Go back to the jog tab and click on the "Jog Increment" button. I see it cycle through 1 0.1 0.001 0.0001 once or twice, then 2 0.2 0.002 0.0002 3 0.3 0.03 0.003 0.0003 then back to two passes through "1 0.1 0.001 0.0001 0.0001" then the "2" cycle then the "3" cycle, then the "1" cycle twice, etc. repeating forever.
(4) Exit Mach and re-start. The jog increment that is shown as the starting jog increment is NOT the one that we ended with before we exited Mach4, and is not even a value from the same location in the 10 element "jog increment" table. I *think* (but don't remember exactly) that is was one of the "1 0.1 0.01 0.001 0.0001" values, but I'm not sure. Clicking on the "Jog Increment" button it changes to one of the new table entries (i.e. a "2" value or "3" value), and then continues to cycle through the "2" and "3" values in order.
So, it is not entirely broken. Just not exactly behaving as expected. The answer seems to be "set the new jog increment table then restart Mach4". I cannot explain why in step (3) above Mach4 seemed to have 2 complete 10-entry jog increment tables - the old one *and* the new one.