[...]Other than looking for a "better" P-channel FET, I don't know what else could be done here.
For the N-channel driver, though, you could replace it with an npn transistor. A common MMBT3904, for instance, has an off-leakage of 50nA.
That's right, a BJT transistor can be a much better OFF switch, especially in higher temperatures. But the BJT VCE(sat) is not a very controlled parameter, and you often have very high VCE(sat) for low-leakage transistors. A MMBT2222 has a very low leakage of 1nA at 30°C, but has a non-controlled VCE(sat) of ~300mV. The PNP 2N2907 has even higher VCE(sat) of -400mV. Those values will vary from part to part, so unless you do calibration, you will have a error voltage that can be significant.
ZETEX specify -1µA MAX leakage for the p-channel ZXM61P02F series, but offer no temperature derating info, so I am suspicious about this. I bet they are as leaky as any other.
To get the utmost lower current, a JFET woud be necessary, with cutoff currents in the picoamp range. The caveats of a JFET are: it is a depletion FET, so you have to maintain a VGS more negative than any of its terminals to have it turned OFF, and its RDS(on) is a bit high, at the hundreds of ohms.
If leakage currents under 1µA and simple circuitry are required, I would use an opto-FET. There are opto-FETs with remarkably low leakage current, like 1nA at 30°C, and rated maximum 5nA at ambient temp. The NEC PS7205B-1A
is such a opto-FET. You can place the FET switch as a high-side switch. The caveat is that you have to drive the LED with 2mA to get a sample, but this can be done at a small duty cycle and sampling frequency, to keep overall current under the required budget. If you only sample at 0.5Hz to 0.1Hz, this can be met.