Be sure to compare apples to apples. If one set of data is with a full load, running against the tide with a head wind and the other is unloaded, with the tide and downwind (ferry boat), you going to see a difference. Same thing if its a genset and one set of readings is full load in the summer and the next is unloaded in the winter. I wrote (three times) the first diesel generator trend analysis instruction for a navy ship back in the early '80's and we found that little things like using a different electric motor for the load or having an entrance door open one time and closed the next made a big difference.