tapin/tapout vs. delay: am I imagining this?

Yeah, it was always my understanding that tapout could not have a delay smaller than the signal vector size, while delay~ could have a delay smaller than the vector size. Unless I'm totally misunderstanding--always a great possibility--it looks as if the vector size is somehow being automatically subtracted from tapout.

BTW, the best description of the difference between tapin/tapout and delay is in the delay~ reference documentation. In the Max 6 reference, you have to click the "Discussion" arrow to reveal the following:

The differences between delay~ and tapin~/ tapout~ are as follows: First, delay times with delay~ are specified in terms of samples rather than milliseconds, so they will change duration if the sampling rate changes. Second, the delay~ object can reliably delay a signal a number of samples that is less than a vector size. Finally, unlike tapin~ and tapout~, you cannot feed the output of delay~ back to its input. If you wish to use feedback with short delays, consider using the comb~ object. Note: While the delay~ object lets you specify time in any of Max's standard time formats, the interval attribute argument should be used when specifying time in any other time unit besides milliseconds or notevalues (e.g. delay 22050 @interval 11025 samples).