When using the alternative rolling method Method VI to have a PC's ability score exceed 18, allocate die roll results to that ability score until the score is 18, keeping in mind that "[y]ou can add as many dice as you want to any ability score, but no ability score can exceed 18 points [therefore i]f you cannot make an 18 by exact count of the dice, you cannot have an 18 score" (Player's Handbook (1995) 19). (For the benefit of readers new to Method VI, a player gets to distribute but 7d6 among his PC's ability scores, and each score starts at 8.)
For races in the Player's Handbook, after ability scores are determined then racial ability adjustments are applied. Table 7: Racial Ability Requirements (27) shows the unadjusted-by-racial-ability-adjustments ability score minimums and maximums; if a character's ability score exceeds or falls below these due to racial ability score adjustments, says the Player's Handbook, that's okay and the ability score that's higher than a maximum or lower than a minimum should be kept anyway.
By contrast, that's not okay for the races in the Complete Book of Humanoids. While it even mentions using Method VI on Generating Ability Scores (7–8), it diverges with this: "Each humanoid race has its own set of ability score minimums and maximums. No ability score can fall below the minimum or rise above the maximum without magical or divine aid" (8). Still, though, racial ability score adjustments are applied after a player picks a race for his PC! Humanoids only implies—rather than addresses directly—how this plays out around the gaming table during character creation.
For example, a player using Method VI allocates dice sufficient to give his PC a Strength 18. Then the player picks as his PC's race the alaghi (18–19) that possesses a +2 racial adjustment to Strength… but, according to its Ability Score Range table, an alaghi possesses a Strength 19 maximum therefore racial ability adjustments would see the PC have Strength 20. Is the player forbidden from playing an alaghi? It seems so: "If you have your heart set on playing a particular humanoid race and you roll ability scores that are too high or too low, the DM has the option to allow you to use the racial minimum or maximum for that ability score" (8 and emphases mine). (Note that here the game blithely assumes players aren't picking their PCs' ability scores as they do with Method VI!) In short, if the player wants an alaghi PC the player must either have the PC be Strength 17 or less or rely on DM largess to lower the PC's racially-adjusted Strength score!
Also of interest in Humanoids is the beastman (19–20), advanced bullywug (22–4), and satyr (53–4) that despite not seeing upward ability score adjustments to Strength nonetheless have maximum exceptional Strength scores listed in their Ability Score Range tables. (Those are, if the reader's curious, 18/50, 18/75, and 18/75, respectively.) How to square these maximum exceptional Strength scores with what Humanoids has offered previously is unclear.
To be honest, I'm not much for Complete Book of Humanoids or Method VI, but if that supplement and that method were used in this DM's campaign, this DM would have players roll d% alongside their 7d6 and use that d% result as their PCs' exceptional score… if the PC's Strength score rises to or is 18 and if the PC's class is warrior.
This makes it so that a player can for his PC pick ability scores then pick race then pick class—the order the Player's Handbook recommends creating characters—and not have to backtrack and pick a different race after rolling exceptional Strength because the player's roll for his PC's exceptional Strength score is too high! The player enters the race selection process for his PC knowing beforehand that were his Strength 18 PC to pick the warrior class his PC's Strength may be too exceptional for some races.
The Complete Book of Elves continues with this change from the Player's Handbook in its section on Ability Scores that says that each race lists the "minimum and maximum scores characters of a subrace may have after the initial [i.e. racial] adjustments have been made. A character who does not meet the minimum requirements of a subrace cannot be a member of that subrace. Scores over the maximum must be reduced" (77 and emphsis mine). Again, no reason is given for this deviation. The weirdness of a PC picking a race then that PC not qualifying to pick that race because the PC's race adjusted the PC's ability scores therefore forcing the PC to pick a different race is not lost on me—apparently one can play a Schrödinger elf.
Anyway, Table 9: Sylvan Elf Ability Scores (81) lists a sylvan elf's Strength as mimimum 6 and maximum 19 with no accounting for exceptional Strength scores. Apparently, a PC who has Strength 18 going in and meets a sylvan elf's other racial ability requirements see its Strength score become 19, no questions asked.
Note: Complete Book of Elves (1992) is so widely despised by fans—for, among other reasons, making elves too powerful—that a 2013 Kickstarter project included as one of its stretch goals Elves author Colin McComb issuing an apology for the Complete Book of Elves. That goal was reached, and that apology is available here.