Ditto Boyd's comments.
Do not take what a map author says about contour intervals as of much value - most do not have a background in cartography and the terms being used have not been defined here.
USGS has used scans of the 24K topos to initially make 1 arc sec (30meter) gridded elevation data and more recently 1/3 arc sec (10meter). LIDAR data is usually used to make the 1/9 arc sec(3meter) data. The process to create these has also become more sofisticated. Therefore, not all NED/DEM data is of equal quality.
On the GPSr the scale changes as you zoom-in and -out. The full set of contours (minor, intermediate and major) are usually set to display at the 800' zoom level. I have noticed on my Oregon 300 that they display at the 1.2mi zoom level - perhaps due to the Oregon's screen having a higher pixel resolution. As you zoom-out, the minor contours will not display; then only the major contours. As you zoom-in past where they are defined with maximum resolution (limited by the GPSr to about the 500' zoom), the lines are simply porportianetly seperated without adding any detail. As USGS selects the contour interval for any area for visibility and a good detail description of the land features, it is up to the mapauthor to do the same for the GPSr mapset.
We also should take into consideration the contour interval of the source topo quads. Using a 20' interval were the source was 40' only uses an algorthim to make more lines - there is really nothing to support the location of these (other than somewhere between the known lines), nor can they show any details between the source contour lines.
1/9 arc sec data was used for my WV and NC contour overlay mapsets. Later this year this data may also be available for PA & OH. It makes huge data sets and there are issues in 'flat' areas.
The 24K verbage was probably used to indicate an CI closer to the printed 24Ks than Garmin's 100K mapset.
Be aware that for WA, OR & CA there are likely to be many more source 24K topos using a CI of 40' (or 80') than 20'. All those additional fabricated lines will tend to make the display difficult to use at 800' and more zoomed-out. John M also indicated he used the 1 arc sec (30m) data, so only 1/9 of the points with 'known' elevations were used to build the contours. It makes a smaller data set, but then why overcontour?
BTW - USGS has made some 24Ks using a 2' CI in very flat/coastal areas. The scale of the map is still 24K.