MAX2839 DC bias cancellation?

I have a MAX2839 connected to an AD9963 as the AFE. The problem is on the ADC side. From the MAX2839 I/Q +/- outputs, each line has: 68pF in parallel to ground, 100 ohm in series to AD9963 input. 

The result is invariably a pretty substantial DC bias on the I lines, and a small one on the Q lines. The AD9963 has a digital offset correction which thusfar has been sufficient to resolve the problem, but the most recent rev of the board requires at least 34 points of adjustment (which is greater than the maximum 31 possible).

See the attached layout. The traces are 100ohm differential impedance matched.

Is there anything else I should be looking at to fix the DC bias issue?

Thanks!