I have been studying a cascode current mirror for an electronics class, that follows this configuration:
As I increase the output voltage Vd4, for most of the sweep the behavior is as expected, but for higher voltages, the current increases (I assume due to channel length modulation, CLM), but I can't really understand how it does not follow the saturation equations (already taking in account for CLM, and using previous estimates from studying the regular current mirror). The plot of the current can be seen below.
Here is a detail for higher voltages.