It shouldn't make any difference, but the lion kingdom finally realized cold air was being introduced from the top & sucked out of the bottom.  It was like pouring water into a pool while trying to suck air from the deep end.  The fan was supposed to mix the air, but maybe it wasn't effective enough.  Filament must have been mixing more air than the fan, thus making it partially work.  The original design had cold air being pumped into the bottom & drained from the top, but this was reversed when the lid popped off.

The fixed routing has the air intake on the bottom, looping over the top, & then ingesting air from below the container next to the sensor.  This should keep cold air from exiting at night & warm air from exiting during the day.  The pump now sucks from the top with the outlet above the inlet. The inside sensor should rapidly detect a dewpoint rise when the pump runs.  

Ideally, both air holes would be on top with the outlet & inlet below the container.  That wouldn't depend on a seal in the pump.

It should be noted the adafruit pump is a surplus inkjet cleaning pump.  Dave briefly revealed one before throwing it away since he couldn't figure out what it was for.

Whether or not the fan circulates the air, it's required to make the sensor accurate.  There's no other place for the sensor except in direct sunlight.

Another test with no filament & fixed plumbing wasn't very encouraging.  The pump didn't stay on anymore, but we don't see interior dewpoint deviating from exterior dewpoint.  The journey began with water presumably coming out of the container itself, getting pumped out, but the nighttime dewpoint didn't fall any more after the pumping.  It's assumed water moves out of the air & condenses on the container at night.  

The 2nd day, a dry cold front moved through, plummeting exterior dewpoint.  A lot of water got pumped out & the dewpoint dropped on the 2nd night.  On the 3rd & 4th days, interior dewpoint rose.  Obviously, absorption by the filament is a key reason why nighttime dewpoint falls.

The sensor is colder than the inside air at night because the fan turns off to conserve electricity.  That might make the dewpoint appear lower.  It breaks away as soon as the fan turns off.

It's believed contraction of the air at night causes outside air to get ingested.  Less air gets ingested when filament is in the container.  That would make valves a good idea, but valves would make the lid pop off during the day.  The valve would have to open when inside temperature was higher than outside temperature.

It already gains a lot of isolation by having an air tight pump & an intake tube long enough to absorb a lot of expansion.  It's never going to be a very efficient process if it is to be affordable.