A quick answer once you are in the middle of your selection process, is to switch the selection. Depending on your data, missed sections should show up as bright blue letting you know you missed them. Less simply, you can use your selection to create a temporary layer in the TOC, and then examine that for misses in various ways including symbology, but if it is fine export it to a new feature.
But if you are often finding this as a problem, and roads from public sources often have incredibly small sections, gaps, attribute errors, perhaps even overlaps, it might pay to process them ahead of time to make selection, symbology, analysis, etc. easier. Setting up a topology to validate your roads may fix gaps under the tolerance, as would running an Integrate. As noted you could set it up to identify dangles. Running the unsplit tool could get rid of all those little sections that share the same attribute but are unnecessarily split. A planarize would get rid of overlaps and split lines where they meet. You might need to explode features if you have multiparts. Finding small sections with attribute errors is harder, you might try selecting for different attributes and then visually looking for those little blobs of blue, or focusing on segments of very small lengths. If you have ETGeowizards there are several tools, including the pseudo nodes nodes and clean tools, that can fix things.
