Without calculating the heat load it's hard to come up with the "right" solution. With mini-splits it's critical to get the sizing right, looking at both the max capacity at your 99% outside design temp as well as it's min-modulated output at +47F (part of the HSPF testing protocol.) A mini-split that's undersized for even the average wintertime load may only hit a seasonal COP of about 2, since it would be running at it's maximum speed/lowest efficiency most of the time. A mini-split that's ridiculously oversized can spend too much power just idling (usually only a problem for PassiveHouse type loads.)
But optimally sized for the load you can hit the high-2s or even low 3s, as long as you don't screw it up.
Some other ways to screw it up are to install the outdoor unit where it is in the drip-line of the roof edge spending all it's time defrosting or can be clobbered by roof avalanches.
A subtler but still real issue in a space heating environment is installing the interior unit high on the wall. Mini-split heads pull in air from the top, and send it out the front out the front. If installed within inches of the ceiling it's incoming air temp is higher than the average room temp, leading to lower efficiency, since then has to cycly up to an even higher output temperature to deliver the heat. Putting it close to the ceiling in a corner sometimes causes it to short-cycle, with bad effects on efficiency. In general it's better to put them no higher than head-height, and it's not insane to put them at chair-height, where the average temp might be close to or below the mean air temp in the room:
If I had to guess I'd hazard that MOST mini-split heads are installed too high on the wall for optimal heating efficiency. If installing it low would be too much of an intrusion into the room, a mini-ducted version would likely beat it on real-world efficiency, pulling in 65-67F air at the floor level rather than 70-74F air with a wall blob mounted at the minimum clearance below the ceiling. HSPF testing presumes 65F incoming air, and that's simply not realistic if pulling air from a few inches under the ceiling.
The Fujitsu xxRLS3H series have a specified output at -15F, and won't stop even if it hits -30F. The drawback with that series is that the min-mod output @ 47F is 3100BTU/hr, even for the 3/4 tonner, so for very small loads it might not be a good fit. The Mitsubishi -FH09NA doesn't put out quite as much heat at low temp as the -9RLS3H, but it has a whopping 11:1 turn-down ratio, and can pull back to ~1500BTU/hr @ +47F, which means it spends a lot less time cycling on/off, but may be maxed out much of the time, depending on your actual heat load numbers.
Bottom line, you CAN make them work at low temps, and you can even make them work EFFICIENTLY, but you can't just throw any cold-climate mini-split at it and expect it to hit the efficiency numbers. You have to calculate the load at both the 99% design temp and at at the shoulder season temps to find the best compromise. Do the math, be aggressive on your assumptions to avoid oversizing, and be smart about where/how you place both the interior & exterior units. A modest amount of oversizing is good for efficiency, but if it's cycling on/off whenever it's above freezing, you've shot yourself in the foot on both comfort & efficiency.