I agree with Melanie.

Let's forget about what the overpaid engineers and datasheets have said, and look at it from purely a results perspective.

With 10-bit A/D the resulting values go from 0 - 1023.
If Vref=5.00V and you apply 5.00 volts to the A/D you get a reading of 1023.

If you then use 1024 as the Vref divisor, the formula to calc the Voltage from the A/D reading would be ...

5.00/1024 * 1023 = 4.99 ; Not the 5V we were looking for.

But if you use 1023 ...

5.00/1023 * 1023 = 5.00 ; this is what I want it to read.
  Notice how the (/1023 * 1023) cancels out.

In the context of this thread, the results are similar.
If the manufacturer states a pressure of 115kPa at 4.7V ... using 1024 with a 4.7V input, you will only read 114.xxkPa, again not the reading we wanted.

So for this (and any other analog projects), I'll continue using the maximum ADC value for the divisor (255, 1023, 4095).
<br>