Thank you for your feedback.
We are surprised by the result as well, can you please send us your LSAPI PHP and FCGI PHP configuration, output of phpinfo().
LSAPI's advantage over FCGI, is the communication cost between LSWS and PHP process, the pure execution speed of PHP script should be the same. So, I expect the result of speed measurement within PHP should be identical.
One possible explanation is that FCGI PHP cache the whole html page generated, and send the script at the end of the request, LSAPI does not use a buffer that big and send partial content, which add some delay during IPC I/O, those delays is added into the overall PHP execution time.
How big is the page generated in your test?
Do you have a isolated test server that you can use 'ab' to measure the performance, not just the speed measurement within PHP?
We will do some tests in our lab.