Oh, interesting. BTW, did you actually mean "specular lights" here or
something else?
FWIW it looks like I used D3DLIGHT and somehow got a fully white quad
out of the directional light instead. Not sure where I tested it but
it was most likely Windows 7 Nvidia. Still, it might explain at least
some of the weird results I got.

Sorry, that's typo, actually 2 of them :(. First of all I meant point and spot lights, not "specular". I was getting zeroes when using them whatever I tried, I tried a lot of states and coordinate tweaking. It was not a background color, it did draw the triangles.

Regarding directional light, I was getting fully white too initially when I was using the test values from ddraw2+ tests. But later after tweaking some values (setting color values to 0.1 instead of 1.0) I found that it actually can output something else, and I matched that something else to the lighting computation with _LOCALVIEWER set to off (not "on" as my buggy comment). I initially suspected that ddraw1 assumes _LOCALVIEWER should be off, but later testing with point lights and using D3DLIGHT2 structure made directional lights in a supposed way, same as ddraw2.

Now I suppose that it is likely some bug in Windows ddraw1, e. g. maybe it is using the flags which are not in D3DLIGHT and gets mad from some random values (e. g., there is a documented flag not to set specular lighting with the light, or not to enable it at all).