NavList:

Ed Popko, you wrote:
"An hour later at 17:05:00, it has substantially jumped to an impossible 336.3 degrees..."

Impossible? No. Let's do it in our heads. You're at latitude 20°N and longitude 74° W. June 24 is just a few days after the June solstice so the Declination of the Sun would be very close to 23.4° N. We shouldn't need an almanac for that! In this longitude we're about five hours from Greenwich. The Mean Sun would cross this meridian (in other words it would be close to local apparent noon) about five hours after Greenwich noon (actually 4h 56m at 74° from Greenwich, but then the Sun is a few minutes slow to the meridian in late June --equation of time-- so we can call it five hours exactly, for the sake of an estimate). At local noon, quite close to 1700 UT, the Sun would pass due north of the observer, azimuth 360°/0°, at an altitude of about 86.6°. Five minutes later, as in your example, the Sun would have moved rather quickly into the northwest (quickly because of the very high altitude), consistent with the azimuth shown in the web app, and its altitude will have dropped a bit. So, no, not impossible! Surprising? Perhaps. But I would bet this is a case where you changed an entry in your experiment, maybe the latitude?, and simply didn't realize it when you looked at the outcome.

You could also test against the new rendition of the USNO web app: https://aa.usno.navy.mil/data/celnav. A minor issue to notice: it lists a date entry format of "YYYY-MM-DD" which is obviously backwards. Their web app also now allows UTC entry to the nearest millisecond, which is ludicrous on multiple levels. In the output data, they've also made a few harmless (and pointless) formatting changes in recent months, but be aware that the underlying engine still apparently outputs coordinates only to 0.001° precision. That would be fine, but when such coordinates are converted to minutes and tenths, as is the standard for this sort of app, the rounding process occasionally changes the last digit by a tenth of a minutes of arc. Therefore the official USNO web app output regularly differs from the official Nautical Almanac as published by 0.1' (my web app does not have this issue). This is insignificant for navigation, but it shouldn't happen in any case.

Frank Reed
Clockwork Mapping / ReedNavigation.com
Conanicut Island USA