You do not specifiy the hardware you are using. If the DAQ device does not perform simultaneous sampling on the two channels, the response is measured at at a slightly different time than the stimulus. This could be part of the problem as the response may not be well correlated with the (random) stimulus. For testing purposes try sweeping the frequency using a continuous sine wave for the stimulus. Make the sweep slow enough that the system response stabilizes at each frequency before movign to the next one. I think the FRF VI will be happier with the sinusoidal stimulation.
On another topic, your diagram is hard to follow. Look at the Style Guide for suggestions. In particular it is preferred that signals flow from left to right rather than zig-zagging all over the diagram. What is accomplished by multiplying by 1?
With a 1 Hz cutoff on the filter and 0.1 second duration samples the filter may not fully settle to a steady state value. If the filter is to eliminate a DC offset, try subtracting the mean.
Lynn