Those do look frustrating. Fair warning: Wall of text below. Just getting started myself so I’m in information overload mode. 
I’ve got the A250 not A350, do have the enclosure, and have prior experience with really ‘hacky’ machines (my first was a Solidoodle 2 Pro, heavily frankensteined to the point of some custom carriage replacement parts and all). So far I’m only printing with Luban, but did make some alterations to the PLA profile and speeds (205C layer 1 / 65C bed, 200C/60C thereafter). Someone else also already recommended only 1mm retraction vs. 5mm which might also be ‘starving’ your extrusion a bit…
Ye eld beaste of Yore:
I’ve gained pretty good prints right from the start with some caution, and am still in the mode of printing almost solely for printing, so to speak. Some of my hacks might not be necessary for you, as they’re somewhat related to my desk setup. But maybe some of them help at least with initial bed adhesion. I actually have barely made any attempt to clean the bed, just a paper towel to wipe off any obvious fingerprint/oil or dust, and adhesion has been fine.
- First suggestions I’d think in your case is extruder calibration, as others have mentioned, and also whether the thermistor is registering properly on the hot end. If you’re getting inconsistent heating, either too low or too high, it can cause really cruddy extrusion, and almost doesn’t matter what you set the Luban temps at if there’s a lot of noise in the measurement. (Too high can cause underextrusion same as too low - insufficient vicosity - because you’re burning crud up in the nozzle and breaking down the material, gunking up the flow path.) You mentioned you already calibrated - you can check again with the M503 command to get it to reply to see if it even applied. I calibrated while wired (USB), and did the M503 ‘check’ after the M500 ‘save’ command the first time, and it didn’t look like it had. Power off/on again and it still hadn’t. I had to make sure to program, THEN power cycle, THEN check to see for sure it was echoing back correctly as changed. As far as thermistor giving accurate readings, see the posts by others about taking it out and puckying a little thermal goo in with it when you push it back in. (I have NOT done this step myself - yet - as I seem to be getting reasonable prints so far, with only some loop dangling associated with undercooling on overhangs.) Related post:
https://forum.snapmaker.com/t/whats-wrong-with-this-picture-printre-hot-ends/7757/21
- Next, and first print for me, is some sort of filament guide. I am NOT a fan of how the enclosure routes the filament, nor a simple bar for a holder. On my old machine re-routing the filament path and creating a better holder with far less ‘backpull’ was really critical for my success, and I’m working on how to do that here as well. (The old machine wouldn’t just get starved if it couldn’t draw the filament well enough, it would actually lose Y-steps because of tension and the print would gradually lean backward as it deposited layers.) In the meantime - I never print unattended. It’s boring to babysit just to nudge the reel or supply slack every now and again, but I think it matters, a lot, until you’re dialed in. Regarding the guides there’s a couple posted on Thingiverse already, I designed my own (may still redesign the loop a bit ‘longer’).
Regarding the spool itself, I had to install the holder ‘upside down’ because I have a shelf over the enclosure. I am working on a better clamp mechanism so I can mount additional reels to my shelf support (1.66in OD aluminum SCH40 pipe), but in the meantime I found two rollerblade bearings I had just barely slid over the existing rod. These help the spool roll a little easier, but I’m still watching it like a nanny until I can get the arrangement I really like, which is a spool that spins absolutely easily. (Yes, that can lead to too much popping off the spool if undampened, which is why the guide and grommet’s ‘friction’ are important as moderators. I’d rather have the spool almost free-running and add a wiper for friction if I had to than that it has to get ‘tugged’ all the time.) If I don’t nudge along right now, even with the skate bearings, I can see where the filament-in-waiting that’s drawn past the grommet is a little ‘bent’ by tension at times between my guide and the grommet itself. To me that’s risk I want to get out of my process.
- Third tip: CALIBRATE THE BED WHILE HOT. I was used to this on my old Solidoodle with a borosilicate glass plate over a really skanky laser cut plywood bed that was warped as crap, and had NO bed autoleveling correction. I’d heat the bed to temp, then heat the nozzle to like 95C - not hot enough I was either leaking or ‘burning’ stuff since I was not extruding, but enough that at least some of the thermal expansion warping had kicked in. Then I’d turn my thumscrews with the paper shim test in a few places on the bed manually, get the best compromise I could for my planned print position, and be off with it. I am still paranoid enough I also auto-level the A250 before every print.
My sequence with the A250 so far has been:
–Manual control to home, then set bed to 65C and nozzle to 95C.
–Wait for both to hit temp then run the 4x4 calibration.
----(I use 2 layers of standard paper and go pretty ‘tight’ on my calf final step, good resistance on pull and immediate paper ruck-up on push.)
–Reset bed temp and set extruder temp to my desired 200C. Wait for heat, ‘load’ filament (basically drool some out to prime it well).
–Then wipe back to File, load gcode from USB, and intiate print. The machine homes and does do a ‘load’ priming before it starts the skirt outline.
Example of a print not intended for the printer, this is an 8020 extrusion ‘foot’ for my mockpit chair setup, printed on a slight remix of the ‘Fast’ settings so 0.24mm layers. I bumped the brightness because potato cellphone pic of black filament, but hopefully you can see it’s pretty clean. (The ‘hair’ on the first shot is truly hair…dang cat…)
