I just spent a while puzzling over a problem until it dawned on me:
a1 = [3.98994e-05 +1i*-0, 4.21732e-05 +1i*3.69849e-05];
Computes to :
0.3990 + 0.0000i 0.0000 + 0.0000i 0.4217 + 0.0000i 0.0000 + 0.3698i
To get what I expected, I had to do the following (strip spaces):
0.3990 + 0.0000i 0.4217 + 0.3698i
I am debugging a large vector written from C code, so my hunch was to spend quite some time looking for bugs in C-code before starting to look at the numerical interpretation in MATLAB.