I certainly do find myself "skimming" code in larger code bases but then I'm also not speed-reading but simply looking at the visual structure (which is where consistent formatting matters) and looking at words in fixed locations (e.g. names in function definitions).
When you actually have to read code (rather than just skim it by looking at structure and words in fixed locations) that usually means you need to focus and carefully follow each expression -- the bottleneck in this case is comprehension, not reading speed (unless you have to keep track of each character, such as when dealing with short and confusing variable names or intentionally obfuscated code).
I was subvocalizing things to remember to do before learning to use task tracking software for that.
You might want to benchmark yourself e.g. with http://www.readingsoft.com/ (I ranked 398 wpm and I know my bottleneck is eye movements)
Also I think code reading speed is only a part of a larger process (e.g. code review or making a fix). For both of these external tools are available making reading speed not as important (iterative development / unit tests for fixing + static code analysis tools for reviews)