I tried simulating what was going on, but everything was behaving correctly, so that means the issue is probably with my direct input code, as opposed to my interface code. Here's a version that generates a binary file whenever you bind an axis. The binary file just contains a binary list of objects for the relevant device and a copy of the state of all buttons at the moment of binding. It also outputs some data when updating the list of bound controls.
Replace your dll, delete/remove/rename LiliPad.ini so I don't have to navigate through the output from loading your button bindings, and try binding analog axes twice. Use different physical axes on your pad each time. Then try binding the d-pad, and post the resulting debug.bin file. For consistency, use the Saitek controller.
Also, when you try and bind the d-pad, does it recognize any axis as being active? That is, if you try and bind the d-pad, then move an analog axis around, does it (try to) bind the analog axis or not? If it doesn't even recognize the d-pad as moving, then when you try and bind the d-pad hold the d-pad down in any direction, and then move an analog axis, so output is added to the debug file.
<Attachment superseded>