PPS and GPSD:

Some GPS receivers offer a 1PPS (one pulse per second) output that GPSD is capable of passing along to ntpd for use as an extremely accurate clock source. These are marked as having PPS in the table above. The "Claimed Accuracy" figures are taken from manufacturers' data sheets, and have not been independently verified.

Where specified, NTP time offsets have been measured by comparison with the NTP time service and reported by gpsd.

Note that either a direct RS-232 connection or the "Magic Macx mod" (1PPS on the GPS chip connected to DCD on the serial-to-USB adaptor) is required for PPS. Typical gpsd/PPS users report ntp clock stability on the order of 1μs (microsecond) for RS232, 1ms (millisecond) for USB. From GPSes that are native USB but not PPS-capable, GPSD can extract time information from the serial data accurate to about 250ms.

Raw Measurements and GPSD:

All gpses require "raw" measurements internally to compute a navigation solution. The ability for users to postprocess these measurements is then a function of firmware and protocol support.

 
Receiver Raw Notes
NavSync No I asked them about raw measurements and they told me that a custom firmware build ($$$!) would be required, and that I probably wasn't authorized to get that data.
SiRF Partial sirfstar1 has raw data in message 5, sirfstar2 has raw data in message 28, sirfstar3 has some raw data in message 28 but carrier phase is not available. rumor has it that carrier phase is available as a premium feature in superstar3.
u-blox Partial The RXM-RAW message contains raw data. This is implemented in all Antaris chips. This is only implemented in the timing versions of Antaris4. In ublox5 it is "only available with premium feature raw data".
Novatel SuperStarII Partial Message #23 contains raw measurements in carrier-phase-enabled firmwares.
Fastrax Yes The "PSEUDO" message contains raw measurements
Trimble Yes Packet 0x5a contains raw measurements
Thales AC12 Yes The PASHR,MCA message contains raw measurements.