Every base-band data signal has two components. One is the data itself (the high and low bit states), and the other is the clock, which is used to position the bit transitions in time. The performance of any digital mode demodulator depends on recovering both of these signals from the incoming signal + noise. If the clock recovery algorithm is good, the data recovery algorithm can be near-optimal in its ability to operate in low SNR conditions. If the clock recovery algorithm is poor, the SNR performance of the data recovery will also be poor, because it will be looking for bit state in the wrong locations. This is something that every modem designer has to deal with. But the JT-* modes and its various brethren don't do clock recovery -- at all. They depend on receiving clock data directly from the system clock, which must be disciplined to a government station, either via NTP, or hand-synchronization through WWV, CHU, or some other means that is not the radio channel from which the data is being taken. In fact, K1JT called this out explicitly in one of his papers on the subject -- he didn't implement clock recovery. The clocks of the end stations have to be pre-synchronized through one of these methods. So whereas demodulators for Olivia, RTTY, PSK, and pretty much every other data mode including CW do the entire task of recovering the signal from the HF channel, the JT modes only recover half (the data), and they must get the other half (the clock) from a non-ham communications channel. As a result, there is no way to meaningfully compare the SNR performance of JT65 with any other digital mode, because the JT65 SNR reports are based on the assumption that the JT65 demodulator has perfect out-of-band clock transition information, which artificially lowers the SNR threshold by using the Internet or other means to transfer half the channel data. Go read the papers K1JT wrote on the subject. It's very clear. He worked around one of the hardest problems in modem design by simply not solving it at all, and taking that part of the transmission from another communications mode. You raise a good point, which is disconnected operation. Technically, this changes nothing, because in order for your modem to work properly, you had to pre-synchronize your local system clock to an NTP, WWV, CHU or other government-sourced clock signal prior to operation, and that one-time act means that you and the other station used a non-ham communication channel to bring your modem clocks into alignment prior to the QSO. Whether your clocks were synchronized during or before the QSO is irrelevant. 100% of the modem clock information was communicated outside the HF channel, and the modem cannot function without this information being communicated in this way. Every other digital mode can recover the clock information from the received tones on the HF channel, and their minimum SNR thresholds are dependent (among other things) on the quality of the clock recovery algorithms they use. JT-65 and its siblings cannot, nor do they even try, to do clock recovery, so the reported SNR numbers are artificially optimistic by comparison. And this is why you can't compare JT65 or FT8 SNR threshold values to Olivia. It is at least as different as apples and oranges. It's like comparing an apple to half (at most) of an orange.