Sliding Window Filter using FIR filter in Simulink

I'm trying to create a "sliding window" or "moving average" filter. Before implementing into my larger model, I created a test model. I have a sine wave input followed by a quantizer, then a discrete FIR filter. The intent is to output the average of the last 20 samples.

The parameters on the FIR filter are:

Coefficient source: Dialog parameters
Filter structure:   Direct form
Coefficients:       ones(1,20)/20
Input processing:   Elements as channels (sample based)
Initial states:     0
Sample time:        .01

Here is the quantized data (left) and the FIR output (right):

0  0

22  8.80000000000000

44  24.2000000000000

44  44.0000000000000

66  58.3000000000000

66  66.0000000000000

88  81.4000000000000

88  88.0000000000000

88  88.0000000000000

88  88.0000000000000

88  88.0000000000000

66  73.7000000000000

66  66.0000000000000

44  50.6000000000000

22  38.5000000000000

22  22.0000000000000

0  1.10000000000000

-22  -15.4000000000000

-44  -30.8000000000000

-44  -44.0000000000000

-66  -64.9000000000000

-66  -66.0000000000000

-88  -88.0000000000000

-88  -88.0000000000000

-88  -88.0000000000000

-88  -88.0000000000000

-88  -88.0000000000000

-66  -67.1000000000000

-66  -66.0000000000000

-44  -44.0000000000000

-22  -31.9000000000000

0  -16.5000000000000

0  0

22  22.0000000000000

44  37.4000000000000

66  49.5000000000000

66  66.0000000000000

88  72.6000000000000

88  88.0000000000000

88  88.0000000000000

88  88.0000000000000

88  88.0000000000000

66  82.5000000000000

66  66.0000000000000

44  59.4000000000000

44  44.0000000000000

22  26.4000000000000

0  9.90000000000000

-22  -5.50000000000000

-22  -22.0000000000000

-44  -44.0000000000000

-66  -56.1000000000000

-66  -66.0000000000000

-88  -78.1000000000000

-88  -88.0000000000000

-88  -88.0000000000000

-88  -88.0000000000000

-88  -88.0000000000000

-66  -77

-66  -66.0000000000000

As you can see, it is not outputting the intended data. It appears to only be considering the current sample and one previous. What have I done wrong?