To make more reliable the comparative test, would be necessary to have the values of measurements corrected (deviations in the troposphere, ionosphere ..) in post processing stages on some server on line of high level, but to do so would have to have the raw values collected before being processed by the gps, and transformed in the standard RINEX format in order to be post-processed. For this would have to make new measurements with the GPS connected to a notebook and program that would collect these data before they are processed by the gps. Sounds crazy, but it's a game very stimulating, I'm learning a lot, and maybe in the end, could at least respond safely the question of accuracy btw the values displayed on the screens of the original and converted 3790, among many other issues that do not fit comment here ..
But I'm slowly walking in these steps, but if anyone has any information on how to extract the data collected from satellite before being processed by the Nuvis / Etrex / Oregons, please let me know. It may be that the .bin format is raw data (gpx is processed), but it is a mere supposition. I have still researching it.
Despite my exaggerated pretensions about accuracy, which go far beyond the needs of the topic, there is no denying that yours proposal of compare tracklogs of the original and modified 3790, with measurements side by side, are ideal for the purpose of comparing the precision "real" BETWEEN the two firmwares. At the first opportunity, I will make that comparison.