The specific answer to your question is to go with the first solution. That is the standard cartographic convention, effectively eliminating the question of where, say, 50.5 goes. As far as where the exact category breaks go (say, precisely 50), it doesn’t matter. Users understand that that the categories are functionally fuzzy. For example, obviously 49.999 is one category while 50.001 is another, but users realize that that the two are practically the same. As far as they’re concerned, “50-ish” pipes, including pipes exactly 50 mm in diameter, can be in either category.
The more broad answer to your question is that if you’re worried about ambiguity of your category divisions, then you need to change your category divisions. Cartography texts recommend that you break your categories at meaningful points for your users –make divisions related to the users’ tasks or decisions. It may not be necessary for the categories to have neat regular intervals. If your users really need to know if a pipe is exactly 50 mm, then maybe that should be a category of its own. User research will help determine the meaningful categories and breakpoints. How do you users group pipe diameters into operationally “similar” versus “different”? I suspect the roughly exponential scale you’re using now is supposed to reflect this, but is it based on actual user research?
For example, you could put the breakpoints at natural places of low data -diameters that characterize relatively few pipes. I would expect there are industry standard pipe diameters, so you make sure no standard diameter is right on the division between two categories. Now there's virtually no ambiguity because there are virtually no pipes to be be ambiguous about. If there are no standards, then find out what diameters exist in the field. Given a statistical distribution of pipe diameters, cartography has mathematical formulas to create categories with minimal variation within each category and maximum variation between categories.
On the other hand maybe pipe diameters gradually grow and shrink over miles of pipe (not very likely, I would think, but it could characterize, say, daily flow volumes, or river sizes). Then maybe your existing divisions are fine. Where is precisely 50? Precisely where the 0-50 code changes to 50-100, implying the diameter goes up in one direction and down in the other. Alternatives to your categories would be equal intervals, nested means, equal progression (e.g., a true exponential, but there are other algorithms too), or categories with equal statistical frequency. Which one you use depends on the results of your user research.
For an introduction to cartography, I recommend the classic Elements of Cartography, by Arthur Robinson et al, now in its 6th addition (1995). I expect there are other excellent texts of more recent vintage that specifically cover electronic maps.