When you first call any method in a package, the entire package must be read in memory. In general, this should be an advantage based on the assumption that if you call one method in a package, it is likely that many related methods will be called either by one procedure that you called, or by subsequent application calls. But this means that the first execution potentially slows down, loading a lot more code may be strictly necessary for a simple function. However, this punishment should disappear as soon as the package is loaded into memory. It does not seem that you are talking about a problem with making the first call to the procedure, although, as a rule, this rules out.
Is there any code in the initialization block of the package that will be executed before performing a small procedure that may distort the results?
How do you determine how long it takes to name these small procedures and what does “much more” mean? Do you call them several times and measure a small number of elapsed milliseconds and see, say, a 30% increase in execution time? Or do you call them thousands of times and see a 1000% increase in runtime?