A big part of the difference you are showing is an unsustainable BABIP (>.330) for Lyles at this point. From a sabermetric standpoint, this tells us that he has probably been unlucky on batted ball hits this year, and that his current results don't reflect his true talent ERA.  A metric like SIERA, which tends to be more predictive, says that his ERA should be more like 4.02.  You may or may not buy this line of reasoning, but I feel fairly certain that the Astros front office takes into account this kind of information.