The loads look a bit strange, specifically 1, 2, and 3 kW?  I thought maybe that should be MW, but you wouldn't run a 1500 kW diesel system against a 3000 kW load ?

The load profile graph you are looking at, with the 1, 2, and 3 kW loads, is the unscaled data (or baseline data as HOMER calls it).  I scaled that data set to average 12 MWh/day, and the scaled peak load is just over 1 MW. 

HOMER's ability to scale hourly data sets (electric load, wind speed, stream flow) is both very powerful and very confusing to newcomers.  In the Primary Load Inputs window, the spreadsheet, the graph, and the stats in the lower left corner all refer to the baseline or unscaled data.  But beneath the graph is a value called the "Scaled annual average".  I set that to 12,000 kWh/day, and since the unscaled average is only 47.8 kWh/day, that means the scaling process increases the magnitude of the load by a factor of about 250.  So the load profile that the system sees is about 250 times larger than the one shown in the graph.  HOMER also adds noise or randomness to the profile so that every day is not the same.  Click the Plot button to see the scaled hourly data.  Here's the graph of the first week, for example:

My habit is to enter the shape of the load profile without worrying about the magnitude.  Then I set the magnitude to what I want with the scaled annual average.