Sorry Steve - I missed the need to respond here - apologies
I cannot say that what we do is Right or Wrong here and am open to input as to what people expect that I would feed back into development, and I could of course question what we do here - however I think the logic here is as follows
If you have two surfaces in my picture below shown as two rectangles. The Larger Rectangle is at Elevation 100 and the smaller rectangle at say 95. There is as you can see an Overlap Zone. The current Merge options are
1) Merge - Maximum - this will create a small surface where the two surfaces overlap only and would be at elevation 100
2) Merge Minimum - this will create a small surface where the two surfaces overlap and would be at Elevation 95
3) Merge - Finish Replaces Existing - in this Case Finish is Surface 2 and Existing is Surface 1 (great terminology in the dialog ....) and you will get a composite surface where the overlap is cut out of the Existing and the Design is dropped in, and then based on Breakline Approximation Parameters as to how much densification you have going on on your linework in the original surfaces you will get varying degrees of steepness of slope and results - note that the Merged Surfaces are always based purely on what the source surfaces look like at the time of the merger and that can generate quite different results as you can see below - the above case is poorly densified data, the lower case is more densified data


We can debate what should happen when you say "Keep Minimum" or "Keep Maximum" - by definition, the two surfaces have to exist at the interpretation point in order to determine which is the Min or Max, in areas where there is only one or other of the two surfaces we could choose to Keep that surface or not I guess, and the developers evidently decided that they should only keep the data where the two surfaces overlap. You could of course argue the other way around and keep a surface everywhere one exists and keep the minimum where the two overlap. I think the thought process here was that if you have an Existing Ground all over the site and you have a Design cut out of it, that when you merge the two using keep Minimum you would get the surface that you have to excavate and if you use the Keep Maximum you would get the surface you have to Fill and if you use the Finish Replaces Existing then you would get the end result surface where the Design is cut into the existing to give an idea of how the site will look once completed.
Of course as soon as you develop a tool for one purpose, people then find uses for it for other purposes and so the tools design intent is now morphed into something somewhat different to the original intent and it can be seen to "n longer work right"
I would be interested to get peoples input on this and how a tool like Merge can be used in different scenarios that may require the tool to be modified in order to better suit the purposes you want to put it to
For me - if two surfaces don't tie correctly before merger, then Merging them is asking for trouble as you will likely never get the right result and I guess that is why I don't use this tool for much other than cutting a Design into Existing - but if we can make it more useful for other tasks please capture them here, provide example data of what you are starting with and what you are trying to achieve and we can for sure look at it.
Based on your question I am assuming that you are suggesting that Keep Min should take the lower of the two surfaces where they overlap, and take whatever surface is there when they do not overlap - so keep Min would look like this

Note the Flags where the two rectangles meet because the two surfaces do not tie together nicely / properly at that location - and without using the source linework to form this surface - you are going to have to explode the TIN to get at the Triangle Edges to fix these issues (merge does not have the linework of the source TINs in the resultant surface)
and if you used Keep Maximum it would look like this

Again - note the Flags and incorrect triangulation because again the two surfaces do not tie together correctly at the outset - so Merging will not fix that and because there are so many variants of data, and people create so many "scenarios" it is extremely hard to predict what is the "right thing to do" in every scenario - however if you have examples of where the tool would work better when you have done the right things to the data - like Offsetting the lines to make the Ties work, or adding sideslopes to tie the surfaces correctly before merging where an overlap scenario is in play - please share and we can discuss.
What you are looking for above is more likely this

and the Merge tool will not solve this as there is a lot of interpretation that is required to make this happen and additional breaklines that have to be created etc.
I hope that this answers the initial question Steve, but I am sure this triggers a lot of discussion which I would love to hear here if possible
We all want the "Big Red Easy Button" for sure ....
Alan