That's an interesting observation and what you would generally expect. Having done a fair bit of testing on the DTP-70 myself, I'd say that increasing patch width improves reliability of measurement for the testing I've done.
Once you start looking at comparing very low dE values there are a lot of variables to consider.
Have the prints had time to fully dry out ? [4 days is a sensible minimum]
Have many times are you measuring each sheet ? [I used 5 sets of measurements, so 10 results to compare]
Does the accuracy of the spectro change with time powered up (heat) ? [possibly]
There are also significant differences between the reliability of reading different paper types and I'd say you can go further and you should assume that different printer models and resolutions will also give other issues with reliability of measurement.
Then a final consideration is does it actually matter when the dE values are so low, especially the errors are measured in very dark values ?
Are the final prints made from the profiles built with such minor improvements in measurement accuracy any better ?
Is this what happens to old spectros?
I doubt it is much to do with the age of spectro, but more modern designs may be better at reading low values.