Problems with DBSRX at GPS frequencies

While trying to pinpoint why my setup with my DBSRX did not work
with the GPS-SDR software ( came across
some funny results. I was using usrp_fft to check the spectrum
in the neigbourghood of the L1 freq when I saw some spikes.

Just to make sure, I removed the antenna, tuned to 1576 MHz, and
played with the decimation range, from 8 to 16. The results can
be seen in this plot:

The plots shown were captured using the average option in usrp_fft,
to make them less noisy.
One can see very clear spikes at different places in every plot.
The only changed variable is the decimation.

Decimation BW range Spikes
d = 8 ± 4.00 Mhz center, ± 4Mhz
d = 10 ± 3.20 Mhz center, ± k * 0.80Mhz
d = 12 ± 2.67 Mhz center, ± k * 1.33Mhz
d = 14 ± 2.37 Mhz center, ± k * 0.57Mhz
d = 16 ± 2.00 Mhz center,

As it happens, I also have a Flex1800 RX-TX borad that covers
the L1 freq, so I repeated my little experiment with the new board:

So this time I did obtain the expected results: a nice flat noise
floor (remember, no antenna), regardless of the decimation factor.

How could these results with the DBSRX board be explained?
Am I missing something basic about the decimation process?
But then, if it is some sort of harmonics related to the
decimation, it should appear in both cases.
Any ideas why this is happening and what could I do to reduce it?
Could this be the cause for my lack of success trying to track a
GPS signal?

Next, I am going to try to record some samples using the Flex1800
and see if they can be postprocessed with gps-sdr. First I have to
find a way to bias the antenna with the flex-1800 since the board
does not seem to have a jumper for that.