A USB 2 interface has no place to connect a PPS signal. If you had a real serial connector on the computer and could somehow get at the internal PPS signal of a USB puck, I guess you could kludge a cable to the serial side of a serial to USB adapter. If you want to solder and build cables, you would be far better off to get one of the evaluation boards or one of the boards made for single board computers (Pi, Arduino, etc.) and build your own enclosure and cable.
My Pi with an older model Ultimate HAT GPS has a two standard deviation accuracy of about 200 microseconds, so yes a real PPS signal does give better results, but the point of my original post was that for $15 you can easily get your time to an order of magnitude better accuracy than that wanted by WSJT. The current Ultimate HAT uses a GNSS chip set so it should be down in the low 10s of microsecond range.
As far as startup times to get the time set, this USB puck will get the time to within the 100 ms that WSJT wants within a couple of minutes of pushing the power button. If one of the first things you do during a remote operation is to turn on the computer, by the time you have your antennas set up and connected, the time accuracy will be in the tens of milliseconds.
Re time2: time2 deals with NMEA sentence timing versus the PPS signal. Since a USB puck has no PPS, it's setting for USB is irrelevant. This is assuming you are running a standard version of ntp and not some variant with different definitions of time1 and time2. Adjusting the time1 value should only be done after the system has been running for several hours at least. You do not have to have any networks servers, i.e. a pool, in ntp.conf to do this.
Re other pucks: Yes, I have found other pucks with other chip sets to have various "issues" which is why I specified the VK-162 G-Mouse which, I have extensively tested specifically for use with ntp. Some of the pucks out there are flat out junk for time keeping purposes.
The VK-162 uses the u-blox 8 chip set which is GNSS, not just GPS, and has 72 channels. I could post a list of all the chip sets I have tested but see no point in it as the VK-162 has been shown to be the best of the cheap ones out there so far and again goes for about $15. If I find a better one, I will post it.
FYI GNSS is the blanket name for all the satellite systems, i.e. GPS, GLONASS, BDS, and Galilleo, so a GNSS chip set has far more satellites it can see.
As for "getting at the bits", the only bit I care about is how well does an out of the box receiver do with ntp.
A Windows machine running Meinberg ntp can be configured to allow it to be used as a ntp server for other systems. Not to downplay the Pi solution, but it would be one less thing to carry to the field that requires power.
I have been toying with the idea of buy a rubidium GNSS disciplined clock which should give ntp an accuracy of a couple of nanoseconds as well as providing a 10 MHz +/- 0.0002 Hz frequency standard. As these things cost about $680, I'll probably toy with idea for a while.