Having worked this out, I should point out a flaw in my question: there is no SetMercator2SP in the python bindings. There is only one SetMercator and you use it to emulate SetMercator1SP or SetMercator2SP by choosing arguments. From what I have seen online, it is possible there are some programming mistakes ... not suprising since this isn't the most common projection.
The function SetMercator has arguments CenterLat, CenterLong, Scale, FalseEasting, FalseNorthing. The argument "CenterLat" seems to be the standard parallel (latitude of unit scale) not an origin, and I believe it is mutually exclusive of Scale. The CenterLong is the longitude of origin. The argument Scale would be ignored if you are emulating the "2SP" arguments, since use of the standard parallel and scale are mutually exclusive.
Note the comment by mdsummer about the only latitude of origin being the equator. My standard parallel was 30 degrees. Since this is substantially above the equator, this means that the data lying right at the standard parallel had a y coordinate significantly greater than zero (zero would be the value of the projection at the equator). However, my data came in this way ... the y coords at the bottom of my grid were appropriately above zero as if the equator were the y origin. Therefore I did not need a FalseNorthing.
Given the ambiguity, in the end I did this directly in proj4:
p=pyproj.Proj(r'+proj=merc +lat_ts=30.0 +lon_0=-128.735 +a=6370997.0 +b=6370997.0')
Note that this use of a and b with these values is identical to "+ellps=sphere" which would shorten the command. However, in my case I had a slightly different earth_radius and I had to sub that value in.