I'll try to show you with screenshots what I'm missing.
In the 1st screenshot, there are polygons with a depth of about 2 m. When adding points with a depth of 0.5 m (shoal or rock), the entire map is recalculated, but this is not correct, there are no other depths as in screenshot 2. I wanted to see the ability to add a polygon without recalculating all the depths nearby, as implemented in GPSMapEdit screenshots 3, 4. Also, the ability to adjust and combine polygons when there is no data as shown in screenshots 5, 6, 7

What is your survey scale (inter-tack distance)?
If a rock isn't visible at the survey scale, mark it with an "Obstruction" point. But if you want everything detailed, use MBES.
5-7 - these are called "Fill Gaps."

Here are data from 16 beam sonar as .csv file.
drive.google.com/file/d/1mAarQV81E_iozbm...FoF/view?usp=sharing

Boening 737 wreck

RM project

RM map

Different software 0.3m grid

Nice data thanks - I'll have a play. We use a 1m grid in v2 with at least some smoothing, that can reduce feature detail.
New interpolation options in v3 should give better results here so it will be useful to use this data as a test.

At Your service, Matt:)
Low cost multibeam sonars is close-future and with Reefmaster is very interesting option.

Very cool datas!

Matt about output At5 files- You're probably know, that newest Simrad and probably future-Lowrance OS not supported this format? RM3 will been compatybile with C-map format which is use in NEON OS from Navico?

@ fjordfisher What is software?

Unfortunately I can't give an answer to this one yet. If they do end up moving exclusively to C-MAP then realistically I think the answer is going to be no, unless they also provide a tool to create C-Map files from some sort of source data - which I suspect they also won't do. We'll see I guess. C-Map is a proprietary encrypted format that they use to ship their map cards so there is little chance they are going to let other companies anywhere near it. In my opinion, of course.

ArcGIS, interpolation with "natural neighbor".
I believe, that in this case is a raster the best solution, and a solution as vector datas in this case is not optimal for RM or other GIS apps.

Good news - we have already implemented natural neighbour and it is coming to RM3. This is an interpolation type that preserves the depth of all data points naturally, whilst still giving results that are (usually) easy on the eye (unlike triangle facet, which is our current only option).

Additional i used "slope" tool to the natural neighbor raster. It's looks a bit like a fake thermal picture.